Home

Visualizza/apri

1. 1 E sem sise Cow s speed d calculation Data parsing in format i i d i m erenses b a d i CowSpeedGraph i l visualization j i I SS Sea Figure 6 37 Flow chart of the algorithms implemented in the software tool for the visual analysis of the cows location data 6 5 SOFTWARE TOOL FOR THE VISUAL ANALYSIS 101 6 5 1 Data selection As described in section 6 2 5 1 during the acquisition phase of the data provided by the RTLS a database was populated The software described in this section was designed in order to perform access to that database and extract the data required for the analysis The user has the possibility to filter data by setting certain parameters in order to perform elaborations on them These elaborations will be provided by the subsequent operations automated by the software In other words it is possible to choose on which cows you want to perform the visual analysis indicating the ID of the tag associated to the cow and which period of acquisition In particular let h be the current time the software requires a number of z hours in order to select the data belonging to the interval h x h Therefore the informations acquired by the RTLS software in the last 2 hours related to a particular cow will be provided Below is the implementation of the
2. LocationEngineCell 0001 Tag 020 000 086 008 Recorded events v Event 9999 aL Clear Slot 45 012 286 020 000 086 008 Delta XYZ 3 32 10 24 1 38 Time 12 24 53 Cell 0001 ra aoa Omoa None Power PPO UWB code Leave trail 1 55212 9 42557 Figure 6 30 Another point of view of the reference tag in LEC 6 3 THE MULTI CAMERA VIDEO RECORDING SYSTEM The accuracy of the localisation and identification achieved through the use of the RTLS was assessed by using information from a multi camera video recording system that was installed in the area of the barn under study Fig ure 6 31 Among possible views obtainable from a video recording system those providing plan views of the barn are the most suitable for the recogni tion of cow behavioural activities Plan views of the barn make it possible to distinguish each cow from the others and determine the real position of each cow within the barn To achieve these objectives the video recording system which was designed according to a previously defined methodology Porto et al 2013 provided panoramic rectified top view images of the barn to obtain real dimensions of cows physical spaces and equipment The application of this system Porto et al 2013 proved that it is suitable to detect cow lying and feeding behaviours in free stall barns 6 4 RTLS PERFORMANCES 8T Figure 6 31 Hardware scheme of the
3. 10 12 13 14 15 function global cardinality of the dataset in this example is 735 var events 735 var locations new Array events for i 0 icevents i locations i new Array 2 locations Dataset 382 47635 468 78619 383 54645 470 03464 383 86748 464 68414 16 17 18 19 20 21 22 23 24 25 26 27 28 29 6 5 SOFTWARE TOOL FOR THE VISUAL ANALYSIS 105 384 15284 474 45772 366 88856 458 22787 377 05451 481 62739 379 58708 475 70617 global locations locations global events events window Then the software performs the construction of the HeatMap by reading the contents of the file containing the dataset dataset js and by using an open source JavaScript library called heatmap js Wied 2011 2013 It uses the power and flexibility of the HTML5 canvas element to draw heatmaps based on the dataset This operation was implemented through the following steps own OO OC fF Q NY Bl N N N BFP Be Be Be hH hH hH HK H e N e O O NO OC Ww NY 55 script type text javascript S rc 7 Stc heatmap js seript script type text javascript src dataset dataset js gt lt script gt script type text javascript gt window onload function var xx h337 create element document getElementBylId
4. 7 fendregion fregion Public Properties public IList Curve Curves get private set public string DataPath get private set public string Description get private set public int Id get private set public bool IsLoaded get private set public bool IsRelativePath get private set public string ThumbnailPath get private set public string Title get private set public string YLabel get private set 7 fendregion fregion Public Methods public void Load load the curves from the XML file XDocument data XDocument Load IsRelativePath HttpContext Current Server MapPath DataPath DataPath Initialize the Curves collection Curves new List Curve 113 Retrieve the plot current plot element from the XML file XElement xplot data Element plot Retrieve the current Description Description xplot Element description Value 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 114 CHAPTER 6 THE CASE STUDY Retrieve the current curves IEnumerable lt XElement gt xcurves xplot Descendants curve Enumerate curves Build a the Curve object and Add it to the current Curves collection foreach XElement xcurve in xcurves Retrive the label string label xcurve Element label Value Initialize
5. Available Sansare Ayalable Events Heb Tags Log 00 11 CE 00 38 87 00 11 CE 00 38 D6 00 11 CE 00 39 F5 M ibi Ele moa Rowe Oppo UWB code ILesvetral 470046349654 Figure 6 16 2D sensors map in the graphical section of the LEC Each sensor is identified by its MAC address represented by a string of 12 hexadecimal digits 70 CHAPTER 6 THE CASE STUDY At the sensor having MAC Address corresponding to that of the Master sensor chosen during hardware installation the Master and the Timing Source flags were attributed 6 2 3 5 System calibration At this point the system was able to identify and detect the presence of a tag placed in proximity to at least one of the sensors but however was not able to accurately determine the relative position in space Therefore we proceeded with the calibration of the system within the study area by using the calibration procedure in order to attribute the missing parameters of localisation to the system The steps that were followed for the calibration of the system were two sensor orientations and cable offsets As already introduced in section 5 3 3 the basic principle of the first phase of calibration also called Orientation calibration once a Ubisense tag was put in a known position within the coverage area of the sensors and the coordinates of the position in the
6. Random positioned pulse train and its frequency spectrum noise like Uie d Lisa x open adero Rey d Frequency spectrum allocation of various wireless systems A Global Positioning System GPS 1 56 1 61 GHz B Personal Communication System PCS 1 85 1 99 GHz C Microwave ovens cordless phones bluetooth IEEE 802 11b 2 4 2 48 GHz D IEEE 802 11a 5 725 5 825 GHz E UWB 3 1 10 6 GHz The bandwidths and power levels of various systems are not drawn to scale Sahinoglu et aL 2008 Definition of AoA between two nodes the reference node black node measures the AoA by determining the angle between itself and the target node gray node ee Signal arrival at ULA having a spacing between the array elements and relation between arrival time differences and AoA Hyperbolic localisation scheme in TDoA measurements Ubisense platform architecture Hardware level contains the Ubisense Sensor Network and optional receivers and sensors Middleware level contains the Ubisense Location Platform which allows the interfacing of RTLS with Corporate Systems and the Applications level that con tains the software applications customized to the use case in question Heathcote 201 D soe ace ee Bee eee we SUA 5 10 Ubisense Series 7000 IP30 Sensors 5 11 Ubisense Series 7000 Compact tag vu 5 12 5
7. DOTTORATO DI RICERCA INTERNAZIONALE in INGEGNERIA AGRARIA XXVI CICLO ANDREA GIUMMARRA UWB BASED REAL TIME LOCATION SYSTEM PERFORMANCES FOR COW IDENTIFICATION AND LOCALISATION AND COW S LOCATION DATA ANALYSIS AND MANAGEMENT IN FREE STALL BARNS TESI PER IL CONSEGUIMENTO DEL TITOLO DI DOTTORE DI RICERCA Tutor Ch mo Prof Claudia Arcidiacono Coordinatore Ch mo Prof Consoli Simona UNIVERSIT DEGLI STUDI DI CATANIA Dipartimento di Gestione dei Sistemi Agroalimentari e Ambientali Sezione Costruzioni e Territorio Catania dicembre 2013 UWB BASED REAL TIME LOCATION SYSTEM PERFORMANCES FOR COW IDENTIFICATION AND LOCALISATION AND GOW S LOCATION DATA ANALYSIS AND MANAGEMENT IN FREE STALL BARNS ANDREA GIUMMARRA DICEMBRE 2013 b 2 Contents List of Figures vii List of Tables xiii I INTRODUCTION 1 1 PREFACE 3 2 OBJECTIVE OF THE STUDY T 3 WORK ORGANIZATION 9 II STATE OF THE ART 11 4 ICT APPLICATIONS 13 Al Anmms ienibosDN uos sess exe Xx e Ey DER 13 4 2 Animal localisation 000002 eee 14 III MATERIALS AND METHODS 17 5 UWB RTLS 19 5 1 ULTRA WIDEBAND SYSTEMS 19 5 11 Ultra Wideband signals lt s e so o o o 21 5111 Signal panditi 2 eds aed m Rs 22 5 1 1 2 Shapeoftheimpulse 29 51 L3 2222222222294 44 23 5 2 LOCALISATION IN ULTRA WIDEBAND SYSTEMS 25 5 2 1 Properties of UWB indoor localisation systems 25 5 2 2 Pe
8. LEDs application controllable Push button application controllable Motion detector US FCC part 15 FCC IDs SEATAG22 SEATAG22HH EU CE Canada RSS Gen RSS 102 RSS 210 RSS 220 Industry Canada ID 8673A TAG22HH Singapore IDA TS SRD IDA TS UWB 3V coin cell CR2477 Industrial adhesive pad supplied industrial Velcro and screw mountings Table 5 2 Ubisense tags technical specification Below there are some typical features of Ubisense tags 5 3 UBISENSE UWB RTLS 37 Precise localisation The tag transmits radio pulses in UWB modal ity which are used by the Ubisense location system to determine the 3D position of the tag within 15 cm By Using UWB technology the precision of the system is maintained even in complex indoor environ ments Two way communication Ubisense tags uses a dual radio system in addition to the one way UWB radio communication used for the spatial detection The tags use a conventional two way communication at 2 45 GHz for the control and telemetry Flexible Update Rate The software platform allows to dynamically change the update rate of the tag depending on the activity If a tag is moving fast we can have a high update for a more accurate localisation but if it moves slowly the update rate can be reduced to conserve battery life When are at rest the tags are put in power save mode via an embedded motion sensor which ensures the restart in the movement events Interactive Bu
9. mented in order to contain the labels of the axes in the graph and return the list of points to be plotted 1 public class Curve 2 region Constructors and Destructors public Curve string label IList lt Point gt points Label label Points points endregion region Public Properties e N e O o AN DA o public string Label get private set public IList lt Point gt Points get private set m endregion Finally the Plot class was implemented order to contain all the data required to build the speed graph such as the Title the ylabel the De scription the Curves etc In addition the Load public method was created to implement the required operations for the preparation of the data to be plotted by using the data structures described above The source code for the class Plot and the Load method is shown below 1 public class Plot 2 region Constructors and Destructors 4 40 41 42 A3 44 45 46 AT 48 49 6 5 SOFTWARE TOOL FOR THE VISUAL ANALYSIS public Plot int id string title string ylabel string dataPath string thumbnailPath bool isRelativePath Id 14 Title title YLabel ylabel DataPath dataPath ThumbnailPath thumbnailPath IsRelativePath isRelativePath Description string Empty Curves new List Curve gt IsLoaded false
10. whereas the other contained the panoramic plan view images recorded by the multi camera system sec tion 6 4 1 2 The information of the two datasets was linked together section 6 4 1 3 6 4 1 1 The dataset of the tags As already described in section 6 2 5 the structure of the records relating to the data acquired by the RTLS stored on disk by the tool specially developed is the following DD MM YYYY HH MM SS ID TAG X Y Z SID ERROR Importing data on database By using Microsoft Office Access software a database was created and it was then populated by entering all the data acquired by the RTLS in a single table This operation has facilitated the manipulation of data In fact formatting operations i e change the time format elimination of duplicate queries and sorting filters by date were carried out Exporting data from database The database was exported to a text file through a tool specially devel oped in C This application initially performs a database connection and then performs a query to the records which are gradually written to a text file The code for database connection and query of select data execution is reported below C A Q N Poe 12 13 14 Select query string strAccessSelect SELECT FROM TagTable Dataset creation DataSet myDataSet new DataSet OleDbConnection myAccessConn null try Database connect
11. 125 UI of the software for storing data acquired by the RTLS The displayed tag 19 apphed toacow in Lii 2 2 eke e Ded 126 User Interface at program startup By clicking on the Open button a window that allows you to search for the folder containing the input of the program is opened som sdo saa ea ess 127 Day Select section of the tool A select box allows you to select the day of stored recording of the two systems for which you want to perform the visual recognition of the position of each tag 128 Software UI for the computation of the true positions of UWB tags 129 Identikit tab of the Software tool It allows the visualization of two images of each cow associated to the identification number of the tag 130 The nine labels each corresponding to one of the tags applied to cows By using these labels the user has the possibility to accurately determine the true position of each tag by placing them appropriately to the desired position by a drag and drop operation 131 LIST OF FIGURES xi 7 9 The computation of the true position of tag with ID 004 The association of a label to each cow made it possible for the operator to place the label 04 on the cow s head by a drag and drop operation The cow true positions shown in the top right section of the UI were automatically stored after clicking on the Save button 2l ll 133 7 10 In Ma
12. 6 return row name 7 8 return unnamed obj DynamicType 9 76 CHAPTER 6 THE CASE STUDY Furthermore the software allows the visualisation of a 2D reproduction of the plan view of the study area which was implemented by using the functions provided by the following libraries of Emgu Computer Vision EmguCV 2013 Importing Emgu libraries using Emgu CV using Emgu CV UI using Emgu CV CvEnum using Emgu CV Structure N In Figure 6 19 the Emgu CV Architecture Overview is shown ZedGraph C Net Presentation C Emu Util gii Net Utilities Emau CV SPU dI GPU Processing C Machine Learning Emgu CV ML namespace C Figure 6 19 Emgu CV Architecture Overview EmguCV 2013 The output produced by the tool is presented in the format shown in Figure 6 20 ANALYSIS AND INSTALLATION OF UBISENSE RTLS T 21 07 2011 11 32 59 160 020020139197 1 234 14 257 4 749 0 0213592443615198 21 07 2011 12 59 52 320 020020140051 10 602 5 942 0 755 0 00154732039663941 21 97 2011 12 59 52 194 020000140053 10 551 9 189 1 53 7 65446202422027 06 21 07 2011 12 59 53 185 020000140051 10 498 5 761 0 894 0 00205906317569315 21 07 2011 12 59 54 50 020000140051 10 59476 177 0 788 0 00124014227185328 21 07 2011 12 59 54 672 020000140053 10 627 7 906 1 699 0 00218551279976964 21 07 2011 12 59 54 915 020000140081 10 59 6 231 1 072 0 00
13. The software saves in a temporary data structure the new posi tions determined by the user to keep any changes in memory Reset All it makes it possible to undo the changes in the current record The software after pressing this button replaces all the labels in the initial position by resetting the changes 132 CHAPTER 7 RESULTS Save it makes it possible to save all the changes The software saves on the disk the information related to the true locations of tags which was placed in the temporary data structure This information consti tutes the output of the software and therefore the dataset of the true positions of UWB tags In the design phase of this software a specific colour was established and associated with some elements of the UI to allow for an easy visual recognition of the position of each tag carried out by the user To this purpose ten distinct colours were associated to the ten tags Subsequently these colours were used to highlight the elements of the UI related to the tags In particular In Tags Data tab near to the coordinates of each tag a graphic element which has the shape of a square and is filled with the colour associated to the related tag was inserted In Identikit tab two images of each cow are associated to the identi fication number of the tag and a graphic element with the shape of a circle coloured according to the specific tag colour In the Map the graphic elements points
14. about 12 25 cm in two dimensions However in field tests accuracy of the Ubisense system was found to vary between 30 cm and 100 cm in the two dimensions x and y in dependence on the applications e g agriculture transit yard management and personnel safety Mok et al 2010 Ward 2010 In this context the main purpose of this thesis was to evaluate the RTLS performances in the identification and localisation of cows during feeding and lying behaviours in the breeding environment considered which included the resting area and the feeding area of the barn To this aim RTLS performances need to be analysed for fixed tags applied to the building structure and for tags in movement applied to the cows A further aim of this research was the reduction of localisation errors through the application of data cleaning techniques and trade off of RTLS identification and localisation performances Then this research study considered the analysis of possible applications of the RTLS to the study of dairy cow behaviour On the basis of these considerations another objective of this study was the development of an automatic and real time software tool for the visual analysis of the cows location data in free stall barns acquired by the RTLS designed in order to have the features needed to be integrated into the world of The developed of a use case of this software tool applied to data acquired 7 8 CHAPTER 2 OBJECTIVE OF THE STUDY b
15. adopted to assess the accuracy of the localisation and identification achieved through the use of the RTLS is described Finally section 6 4 shows the materials and methods relative to the computation of localisation and identification performances of Ubisense UWB RTLS while section 6 5 describes the au tomatic and real time software tool developed for the visual analysis of the cows location data in free stall barns acquired by the RTLS Part IV describes the results achieved in this research study In particular chapter 7 shows results relative to the study of localisation and identification performances of the RTLS section 7 1 and those relative to the software tool for the visual analysis section 7 2 Finally chapter 8 reports in detail the comments on obtained results by making comparisons with other similar works and the possible further improvements of the research in this field Finally part V is dedicated to the outcomes achieved in the research work described in this thesis 10 CHAPTER 3 WORK ORGANIZATION Part STATE OF THE ART Chapter 4 ICT APPLICATIONS FOR ANIMAL IDENTIFICATION AND LOCALISATION INDOOR The automatic detection of changes in dairy cows behaviour combined with the concept of IoT introduced in chapter 1 can be used as input to non invasive warning systems which alert the farmer when some health problems e g lameness or a particular physiological status e g estrus occur In this cont
16. best estimated online in real time 3 a target value and or trajectory for each process output e g a be havioural pattern pollutant emission or growth rate 4 actuators and a model based predictive controller for the process in puts In general the reliability of the PLF is determined mainly by the ani mals and from all the relevant physiological variables that can be measured continuously as the weight the respiratory rate or heart rate body temper ature motor activity behavior the produced noise food intake etc Possible approaches to automatic monitoring systems can be based on sound images location and environmental data collection Information Technology IT is continuously doing significant progress in terms of technical efficiency In particular the new techniques of electronic design reduction of size and power consumption of the devices and the grow ing diffusion of open source software have made the technology affordable and accessible When IT and the sensors are combined with the use of internet it s possible to implement new technologies able to support the farmer by providing him with early warning systems for the automatic detection of potential production health and welfare problems Tullo et al 2013 Today internet is used as a global platform for machines and human communication web chats emails etc and for the interconnection of smart objects As a result it is expected that in the near
17. ends with an arrow send input to sensors send request to send response to sensor send request tag send locaton data Figure 6 6 State diagram A solid black circle represents the starting point of a sequence of states while the point of arrival is represented by two concentric circles in which the innermost is filled with the black color In each state its name is shown in the upper part of the rectangle and activities i e what happens when the system enters a given state in the lower In Figure 6 6 the state diagram of the case study in question is shown By proceeding from the starting point once that the input is sent to the sensor you switch to the Sensors available state in which the activity transmit RF signal occurs From this state you can switch in two other states through the send request to tags transition which leads i to the Tag not available state if the tag does not respond and then consequently at the point of arrival or ii to the Tag available state if the tag answers Subsequently you switch ANALYSIS AND INSTALLATION OF UBISENSE RTLS 59 to the Cow Localised and Identified state through the send response to sensors transition which contains a sub state diagram which represents the implementation of the techniques of positioning adopted by the RTLS We move finally to the Data updated state through the send location data transition 6 2 2 4 Sequence d
18. heatmapArea radius 6 visible true opacity 50 legend DE title Number of Events Distribution var locations window locations var pos for i 0 i lt window events i pos locations xx store addDataPoint pos 0 pos 1 106 CHAPTER 6 THE CASE STUDY The heatmap js library creates a global object called 7h337 This global object has a function create that takes one argument config and returns a heatmap instance In the Javascript programming language each primi tive is represented by an object An object is just a special kind of data with properties and methods Properties are the values associated with an object whereas methods are the actions that can be performed on objects Crockford 2008 In this case the object h337 executes the action to cre ate a heatmap instance invoking the method create and stores the return value in the variable xx 6 5 3 CowSpeedGraph implementation Asit is shown in the flow chart of Figure 6 37 the implementation of CowSpeed Graph such as the CowHeatMap was managed through two separate mod ules the server side and client side Sections 6 5 3 1 and 6 5 3 2 describe the operations performed by the in dividual modules 6 5 3 1 Server side The server side module is designed to provide services to the client side mod ule which forwards to it the requests dictated b
19. the support of a user manual An important feature of this software is that it allows you to perform the analysis for each animal This property is particularly important for the cases where the behavior of a cow must be analysed separately i e not in relation to the group it belongs Martinez Ortiz et al 2013 calculated the speeds of approximately 190 dairy cows inspected by a video tracking system over a number of weeks Moreover the relative speed of a cow with respect to the group on its own is not sufficient to detect lameness a cow may be consistently slower than the group due to old age or simply due to its own preferred pace of walking Therefore they proposed to detect lameness by monitoring each individual cow s speed over a number of days to look for consistent changes in mean speed Another important feature of this software is that since data for all 24 hours for each day are available you do not have the problem of lack of the data which may contain useful information to analyse This type of problem for example is often typical in cases where video tracking systems are used since they can record only during daytime hours The use case carried out in this work showed also that a pattern related to the cow behavior analysed i e when the state of estrus occurred could be identified in both CowHeatMap and CowSpeedGraph In particular if the CowHeatMaps are considered the test showed that in general the cow occupied
20. 08 15 60 21 534850 2011 08 1560 23 534851 2011 08 156025 li n Figure 6 33 Images dataset structure The number of text files is equal to the number of the days when image recordings were carried out In the figure is also reported an example of the contents of one of these text files corresponding to the day August 15 2011 6 4 1 3 Matching between tag dataset and image dataset In this phase the matching between the two datasets built in the previous sections was carried out and in particular between the dates of acquisition of the images and those of the tags in order to intersect the data acquired by both systems and verify the planimetric position of each tag provided by the RTLS Therefore to each record belonging to the dataset of images a record belonging to the dataset of tags was coupled based on the following criteria Each panoramic plan view image panoramic characterised by an ac quisition time was associated to tag positions which were recorded When it was not possible to associate to a panoramic the position of all the tags present in the scene because of differences in acquisition time the excluded tags were even then associated with panoramic by using for them the positions acquired at the time t closest to t within the interval t 3s t 3s This time interval was selected according to the acquisition time rate of the RTLS 6 4 RTLS
21. 12th August 2013 the cow having the tag in question showed the status of estrus which was visually observed by the farmer at around 9 30 am By observing the CowHeatMaps the position of the location points in the space maintained on average a pattern that was repeated nearly in a systematic manner in each of the days selected for the analysis with the exception of 12th August 2013 where a distribution of points in a much wide area was observed In particular since the data were related to the CF Time interval the most frequented zones in the study area by cow in question were mainly located at the manger as expected and in particular at the left passage lane probably due to the presence of the watering hole and the cubicles in the middle or the left on the resting area in particular by referring 7 2 SOFTWARE TOOL FOR THE VISUAL ANALYSIS 147 2013 08 10 06 00 01 COW SPEED 0 09 1000 1208 1490 500 1800 a Tag Id 020 Date b Tag Id 020 Date 10 08 2013 10 08 2013 Interval CFTime Interval CFTime 2008 11 06 00 02 COW SPEED 0 1 0800 10 00 12 00 1400 1600 1800 c Tag Id 020 Date d Tag Id 020 Date 11 08 2013 11 08 2013 Interval CFTime Interval CFTime 2013 08 12 06 00 03 COW SPEED 0 02 10 00 1200 1400 1600 1800 e Tag Id 020 Date f Tag Id 020 Date 12 08 2013 In 12 08 2013 Interval CFTime terval CFTime Figure 7 17 CowHeatMaps and
22. 21 Wied P 2011 2013 Js heatmaps Retrieved 11 10 13 18 13 from http www patrick wied at static heatmap js index html 170 References Yan J 2010 Algorithms for indoor positioning systems using ultra wideband signals Unpublished doctoral dissertation Delft University of Technology Delft The Netherlands Yong X Yinghua L Hongxin Z amp Yeqiu W 2007 An overview of ultra wideband technique application for medial engineering In Com plex medical engineering 2007 p 408 411 IEEE ICME International Conference on IEEE Zhou J amp Shi J 2009 Rfid localization algorithms and applications a review J Intell Manuf 20 695 707
23. 32 187 0 99 0 99 0 11 0 05 Table 7 5 RTLS identification and localisation performances computed for each one of the eight tags analysed and for the reference tag ID 187 for the Metric B system performances obtained by adopting the Metric A and those obtained by adopting the Metric B showed that the best trade off between localisation and identification performances was achieved when Metric B was adopted Precision and sensitivity resulted almost equal in Metric A and Metric B In fact the number of false positives and the number of false negatives were set equal to 0 in Metric A and these numbers resulted in an irrelevant quantity equal to about 1 9 of all the detections performed by the RTLS in the case where Metric B was adopted Furthermore localisation performance was higher in Metric B than in Metric A since the value of mean error standard deviation obtained with the Metric B was lower than that of Metric A Similarly the comparison between the performances obtained by adopting the Metric B and those obtained by adopting the Metric C showed that the best trade off between localisation and identification performances was obtained when Metric B was adopted Although localisation performance obtained with Metric C was better than that obtained with Metric B the number of false positives and the number of false negatives obtained with the Metric C were much higher than those of Metric B since they were equal to about the 39 7 of t
24. CowSpeedGraphs relative to the tag with ID 020 in CF Time interval distinct by day 1 3 148 CHAPTER 7 RESULTS 201308 13 06 06 39 COW SPEED 0 01 0700 0900 1000 1100 1200 1500 1400 1500 1600 1700 1800 1900 Time g Tag Id 020 Date h Tag Id 020 Date 13 08 2013 13 08 2013 Interval CFTime Interval CFTime 2013 08 14 06 00 05 COW SPEED 0 22 0700 08 00 09 0 1090 1140 1200 1300 1400 1590 1600 1700 1800 1900 Time i Tag Id 020 Date j Tag Id 020 Date 14 08 2013 In 14 08 2013 Interval CF terval CFTime Time os 2013108 15 0609 25 COW SPEED 0 12 O omw 0800 0000 1020 1100 1200 1500 1400 1500 1600 1700 1800 1900 Tine k Tag Id 020 Date 1 Tag Id 020 Date 15 08 2013 In 15 08 2013 Interval CFTime terval CFTime Figure 7 17 CowHeatMaps and CowSpeedGraphs relative to the tag with ID 020 in CFTime interval distinct by day 2 3 7 2 SOFTWARE TOOL FOR THE VISUAL ANALYSIS 149 2013 08 16 06 00 06 COW SPEED 009 Cow s Speed m s p amp Bow x Number of Location Events 1 0700 osoo 0900 1000 11 1200 1300 1400 1800 1500 1700 1800 1900 Time m Tag Id 020 Date Tag Id 020 Date 16 08 2013 In 16 08 2013 Interval CFTime terval CFTime Figure 7 17 CowHeatMaps and CowSpeedGraphs relative to the tag with ID 020 in CF Time interval distinct by day 3 3 the Figure 6 10 On 12th August 2013
25. ORE S 71 O20 Data Sree sche eA gm vom ego 73 6 25 11 Populating Database sa ro x fed DoD Statie tegi e uu dues oec de Ee m Ro wD 79 6x1 Gow tas applicant uuu bed eee ee eee ee ee 80 6 2 6 Filtering alporithms ne teene Ro x 82 DB Cow localisation a 9 X bee See eS 83 6 3 THE MULTI CAMERA VIDEO RECORDING SYSTEM 86 64 RTLS PERFORMANCES s 87 6 4 1 Automatic data preprocessing 87 6 4 1 1 The dataset of the tags 88 6 4 1 3 The dataset of the images 90 6 4 1 3 Matching between tag dataset and image dataset 92 CONTENTS M 6 4 2 The dataset of the true positions of UWB tags 95 6 4 3 Computation of the planimetric localisation errors of thietaps so bk ek we ee ea eee 96 6 4 4 Assessment of the localisation and identification per formances of the RTLS 96 65 SOFTWARE TOOL FOR THE VISUAL ANALYSIS 99 Dol ou us d obe ove TR ede RR 101 6 5 2 CowHeatMap implementation 102 6 5 11 Serverside 24 4424 moo ow lt 102 DS Chen eue uuu sod Romx 104 6 5 8 CowSpeedGraph implementation 106 6 5 9 1 ssc tee 106 0 3 5 2 side cus e gogo Row Rs 116 6 5 4 A use case of thesoftwaretool 117 IV RESULTS AND DISCUSSION 121 7 RESULTS 123 nl BUGS PERFORMANCES ox ES 123 7 1 1 The RTLS operati
26. The table in the figure shows the MAC address of the sensors the cell to which they belong the current status date and time in which they were active before the last update the IP address and any flags respectively ANALYSIS AND INSTALLATION OF UBISENSE RTLS 69 tool in order to precisely mapping the sensors The position of each sensor was inserted manually by bringing the values corresponding to those of the here the sensors were physically installed real reference system w 00 11 CE 00 38 87 Power 00 11 CE 00 38 87 Cumulative 4000 100 2000 5 0 0 o 5 1188 1783 2378 0 1600 2400 3200 4000 172549 00 11 CE 00 38D6 Power 00 11 CE 0036 Cumulative 4000 10 2000 4 5 E 1783 zin 90 CJ 1800 4000 LO 299038 0011 CE 00395F Power 0011000338 Cumulative 7 si 1788 1783 27 CE 10 2400 3200 4000 0011 0E 0039 5 Power 00 11 CE 00 39 F5 Cumulative 1188 1783 237 un 800 1800 2400 300 4000 282353 denotes current threshold Cree Comte Figure 6 15 The power thresholds of each sensor indicated with a red horizontal line allow to distinguish the noise signal green wave from the UWB pulses transmitted by the tags Ed Map Cel Sensor Sensor and Cals Sensor status Eisis E Col Plan LocationEngineCel 0001 D011 CE 0039F5 00 11 CE 00 38 8F 0011 003887 01 0 003806
27. a new Set Retrieve the points from the XML file IEnumerablecXElement xpoints xcurve Descendants point Enumerate the points Parse and add their values to a new Point object Build the Set IFormatProvider culture new CultureInfo fr FR true List lt Point gt set from xpoint in xpoints let x DateTime Parse xpoint Attribute x Value culture let ymin float Math Round float Parse xpoint Symb Value Replace CultureInfo 4 let y float Math Round float Parse xpoint Value Replace CultureInfo ie 4 let ymax float Math Round float Parse xpoint Attribute ymax Value Replace CultureInfo CurrentCulture 4 select new Point x y ymin ymax ToList Add the curve to the Curves collection Curves Add new Curve label set Set data as loaded IsLoaded true 6 5 SOFTWARE TOOL FOR THE VISUAL ANALYSIS 115 84 ss endregion s6 The just reported instructions implement data extraction to be plotted from the Xml file in order to create a file Html to be sent to the client Being a client server system the choice to use a html file for the trans mission of information processed by the server to the client was based on the fact that the typical features of such a data structure guarantee high performance in terms of transmission speed
28. applying an indoor localisation system based on Ultra High Frequency UHF technology Another group of studies Barbari et al 2008 Porto et al 2012 Reiners et al 2009 has been carried out with the aim of demonstrating that RFID technologies based on High Frequency HF and UHF could locate individ ual animals inside specific functional areas of buildings for intensive animal breeding The outcomes of these studies demonstrated that it possible to detect the animals in specific functional areas of the breeding environment e g feeding area and resting area though high localisation errors are found with these systems The localisation accuracy provided by RFID systems based on HF and UHF technologies could be improved by using Ultra Wide Band UWB technology lvarez amp Cintas 2010 The advantage of this technology is provided by the signal transmission mode which takes place by means of short duration pulses This mode of transmission provides UWB based systems with a low sensitivity to interference due to the absence of reflection of the wave itself Furthermore the low radiation used in UWB technology has allowed utilisation of a UWB system for remote monitoring and measuring the patients motion in short distance within hospitals Yong et al 2007 16 CHAPTER 4 ICT APPLICATIONS Part MATERIALS AND METHODS Chapter 5 THE ULTRA WIDEBAND REAL TIME LOCATION SYSTEM As already seen in chapter 4 localisation a
29. carried out only once since the components layout of the structure used for the test remained unaltered during the whole period of observations Otherwise it would have been nec essary to perform further calibrations whenever the layout of the elements present in the area under study had been altered This observation belong to in the recommendations made in the work carried out by Shahi et al 2012 ANALYSIS AND INSTALLATION OF UBISENSE RTLS 1 6 2 4 Data acquisition After having carried out the operations above described the RTLS was able to identify and determine the location of a tag placed at any point belonging to the coverage area of the sensors In this phase the choice of the acquisition time interval of the tags po sition during the test was crucial In fact as mentioned in section 5 2 2 it s important to maintain an optimal trade off between the batteries energy consumption and the update frequency of the measurement A solution to this problem consists in being able to set an update rate related to the ac quisition of the measure which ensures a battery life within the desired time period but respects at the same time the desired update frequency of the measurement On the basis of this concept an acquisition time of 0 865 seconds was chosen in order to ensure a battery life for a period of at least eight weeks for the tags applied to the cows This choice was based on the idea that the average speed of a cow in the
30. checked that the graphic element associated to the tag was displayed in the right position When the operator did not find this condition he moved the graphic element in the correct position and the soft 96 CHAPTER 6 THE CASE STUDY ware recorded a new position of the tag y After a linear conversion from pixels to meters these new positions of each tag for each measurement constituted the dataset of the true positions of the tags during the monitoring period 6 4 3 Computation of the planimetric localisation errors of the tags For each tag a planimetric localisation error was obtained by computing the Euclidean distance between the position provided by the RTLS and that verified by the operator m a l WD G 1 n 61 These errors were utilised to compute localisation error at 90 percentile mean localisation error and related standard deviation Identification and filtering of anomalous measurements which were highly different from the central data distribution values were carried out by adopt ing an outlier data cleaning technique Peck et al 2011 The measurements higher than q3 w 9 qi and lower thang q3 q where qi and qs are the 25 and the 75 percentiles respectively and w 1 5 were discarded 6 4 4 Assessment of the localisation and identi cation performances of the RTLS Localisation performance of the RTLS was evaluated by using mean localisa tion error and standard d
31. data It gives the administrator control over objects building models visualization and geometry Ubisense 20132 La Table mostra alcune specifications della Ubisense location platform Specifications Supported platform Windows XP Vista Business Edition 2003 Server 2008 Server and Windows 7 NET API requires NET 2 0 or higher Visualization tools require DirectX 9 0 or higher Linux 32 bit or 64 bit 2 6 kernels Security Supports AES 256 encryption for all data Open APIs NET 2 0 API to all features as standard C API Windows or Linux also available Table 5 3 Ubisense location platform specifications 5 3 UBISENSE UWB RTLS 41 5 3 1 3 Software applications As mentioned in the previous section Ubisense software platform allows you to create applications interfacing directly to the Ubisense Location Platform and to make calls to all the services that were already described in the previous sections through the use of Ubisense NET APIs In particular it is possible to create console applications Windows Forms applications web service applications etc in different programming lan guages C C Visual Basic etc by using Microsoft R Visual Studio Express a programming environment free distributed by Microsoft R To include reference to Ubisense to be assembled within the project that you want to develop just select the file with dll extension in the folder where the Ubisense software was in
32. devices equipped with temporary sleep in situations in which the measurement is not needed for example when the object to be localised is in a stationary position the device may set itself automatically in a state of stand by until it receives the next movement through appropriate motion sensors in order to optimize the costs due to the consumption of the batteries Power consumption vs update rate An important detail also re lates to the maintenance of an optimum trade off between the energy consumption of the batteries and the update frequency of the measure ment A solution to this problem requires the ability to set an update rate regarding the acquisition of the measurement which ensures the battery life within the desired time period but which respects at the same time the desired update frequency of the measurement The estimation cycle length This feature in particular cases is not to be considered of minor importance Depending on your goals in fact can be a very important factor Just think in certain situations in which you must automate the occurrence of an event as a result of another directly connected For instance when activating an actuator such as the closure of a fence following the movements of the animals located within a livestock environment whose movements are tracked by a localisation system an update cycle rather long may cause delay in the closure of the fence Scalability A good localisation system mu
33. environment in question is less than 0 5 m s and therefore it would have been possible to perform the tracking of the animal without losing important information in any context related to the behavior e g lying standing and feeding The chosen time rate was manually set by the software thanks to the specific functionality offered by the LEC software tool In Figure 6 17 the section of the LEC dedicated to the management of the acquisition time and the real time monitoring of the battery status associated with each tag is shown Location Engine Contig 8 8858588588 Figure 6 17 Battery status and update rate of each tag 72 CHAPTER 6 THE CASE STUDY Since there is only a single UWB channel only one tag can be located at a time The Location Platform divides time up into a number of time slots and allocates appropriate time slots to tags so that their update rate is as close as possible to the rate requested for the tag The conventional radio frequency is used to manage this location schedule to receive beep and flash requests and to send button presses and battery power status information The UWB channel is used only when generating a location at the scheduled time for each tag The tags supports an automatic sleep feature to save battery When a tag is still for a few moment it will automatically enter sleep mode if this feature is activated To wake up
34. error at the 90 percentile of the eight tags resulted equal to about 0 51 m and about 0 97 m respectively The number of points verified for each of the eight tags applied to the cows were lower than those verified for the reference tag Tables 7 2 and 7 3 This was due to the fact that all the localisation data of the reference tag provided by the RTLS were considered since the position of the tag in the barn was known and fixed during the trial Therefore the errors were obtained by computing the Euclidean distance between the point deriving from the localisation of the tag and its known position On the other hand the number of points verified for each one of the eight tags inserted in the cows ears resulted lower because it depended on the assessment of tag position in the panoramic image and the rule of automatic association between the image dataset and tag position dataset In this rule it was established that if the cow was outside the areas monitored by the video cameras i e the feeding alley or the stalls or else the acquisition time of the tag was not within the time interval defined to determine the matching with an image the measure was discarded 7 1 RTLS PERFORMANCES 139 7 1 5 Trade o between localisation and identi cation performances of the RTLS Tables 7 4 7 5 and 7 6 shows the identification and localisation performances of the RTLS relative to Metric A Metric B and Metric C respectively In particular
35. i e to perform the operations necessary to ensure that tags are working properly Once the Activate tags use case is activated a sequence of use cases is triggered In particular this use case is an extension of the Locate tags use case while the tag battery is discharged which represents the main use case diagram In fact it includes Compute TDOA and Compute AOA use cases in order to indicate the two positioning techniques used by the system and extends in Activate sensors use case while the cables are connected and The extend relation allows you to create a new use case by adding more steps to an existing use case The include relation allows you to define one or more use cases within another use case ANALYSIS AND INSTALLATION OF UBISENSE RTLS 55 the DHCP server runs Furthermore it is an extension of Update and store data in order to indicate that once location is carried out data are updated and saved in memory by the system This last use case is an extension of Load software use case Activate sensors User lt extend gt lif battery of tag Compute ccndudln is not discharged ittu TAERE R gt extends gt Compute AOA gt include 7k tif cables are connected and lt lt t S3 extend gt if DHCP server runs xD Figure 6 4 Use case diagram 6 2 2 2 Class diagram Class diagr
36. in Real Time Each sensor is able to sustain a continuous update frequency of 160 Hz which means that the tag can be seen every 6 25 milliseconds from each sensor or cell Flexible and scalable installations The Ubisense system uses cel lular architecture for scalability from small to large installations Thou sands of sensors can be integrated into a single enterprise wide system to monitor an unlimited area and manage thousands of tags Wired or wireless The sensors can be connected with standard Eth ernet cable or via Wi Fi wireless adapters by using the existing infras 5 3 UBISENSE UWB RTLS 35 Sensors technical specification Dimensions 20 cm x 13 cm x 6 cm Weight 650 g Temperature range from 20 C to 60 C Humidity range from 0 to 9596 non condensing Standard operative range gt 160m Precision 2 15cm in 3D UWB channel 6 GHz 8 GHz Telemetry channel Narrow band 2 45 GHz Certifications US FCC part 15 EU CE Power supply Power over Ethernet IEEE 802 3af Low voltage 12V DC 10W optional Mounting options Adjustable mounting bracket Table 5 1 Ubisense sensors technical specification tructure such as access points Ethernet switches and CAT5 cabling for the communication between sensors and servers Two way communication Ubisense sensors support two way com munication at 2 45 GHz with Ubisense tags This allows the system to dynamically manage optimally tags for example to dynamically ch
37. instead the distribution of the points in space mainly involved the manger the service alley the left and right passage lanes and the feeding alley In addition by observing the coloration of these points it was clear that in the CowHeatMap of 12th August 2013 the percentage of red points cover the total number of points represented in the CowHeatMap of that day was significantly lower than the same quantity measured in other days Furthermore in Figure 7 17 on the right of each CowHeatMap the rela tive CowSpeedGraphs are shown where the instantaneous speeds calculated by the software were plotted By observing these graphs also in this case on 12th August 2013 higher values of instantaneous speed over time were recorded especially in the time interval ranging from 8 00 am to 10 00 am compared to those recorded in other days This proves to be a confirma tion that the motor activity manifested by the cow having tag ID 020 in 12th August 2013 was notable compared to that recorded in the other days observed Similarly Figure 7 18 shows CowHeatMaps and CowSpeedGraphs relative to the tag with ID 020 in CLTime interval distinct by day Also in this case the CowHeatMaps show that the distribution of the location point in the space followed a pattern that on average was repeated in all the days chosen for the test with the exception of 12th August 2013 when once again a distribution of points in a much wider area was observed In parti
38. it allows you to set the color of the grid lines Circle size it allows you to set the size of the circle displayed at the pointer location when positioned within the area of the chart Stroke width it allows you to set the thickness of the line representing the plotted curve Show roller it lets you choose whether or not display the roller i e the value of the display scale Rolling Period it allows you to set the value of the roller To view the CowHeatMap relatively to the parameters selected you must click on the element which is associated to a HeatMap icon situated in the controls section Figure 7 14 shows an example of CowHeatMap The CowHeatMap will appear in the visualisation section The number of events distribution varies depending on the dataset of the CowHeatMap and the color varies from red maximum concentration of the location events in the point to transparent null concentration of the location events The mini mum concentration of events location is indicated by the blue color To view the CowSpeedGraph related to the selected parameters you must click on the item which is associated to a SpeedGraph icon in the controls section Figure 7 15 shows an example of CowSpeedGraph As de scribed in section 7 2 1 2 the CowSpeedGraph is interactive In Figure 7 15 in fact you can see the zoom operation performed on a region of the graph through the combined click and drag actions Figure 7 16 shows the view o
39. letter capitalized The list of methods is graphically represented below the list of attributes and separated from this by a horizontal line The methods may have additional informations In the parentheses that follow the name of a method in fact you can show any required parameters to the method together with their type Finally if the method is a function the return type must also be specified In Figure 6 5 Location Engine Configuration class for example is constituted by nine methods including sensorCalibration which represents the calibration function of the system When several classes are conceptually connected with one another such connection is called association Graphically an association is represented with a line connecting the two classes with the name of the association just above it The multiplicity is a special type of association in which you show the number of objects belonging to a class that interacts with the number of objects of the associated class There are different types of association and each of them is represented by a default symbol In Figure 6 5 the class diagram of the case study in question is shown The Sensor class is connected to the Tags class in two ways the first indicates the association that is established when a sensor interrogates a tag and the second indicates the response of the tag The multiplicity of these associa tions is 1 since each sensor interrogates one or more tags and each
40. locally i e in the same vir tual machine where the software and the Ubisense Location Platform were launched Therefore through a SQL query a table is created if it not existing where the cow location data through an insert query can be stored After this operation the information is automatically removed from the data temporary structure Finally after disconnecting from the SQL Server database all the steps described above are repeated until the end of the application 6 2 6 Static test An accurate operation of testing after a careful calibration phase and af ter the development and installation of the software for the storage of data acquired by the RTLS was performed inside the study area This test con sisted in placing a tag in a number of 25 known points and then record the location data acquired by the RTLS for a few minutes This test allowed to evaluate the localisation performance of the system under static conditions along the entire area and determine any malfunctions related to the geometric conditions Another objective of this test was to store the results obtained in order to subsequently compare them with those obtained in a different scenario where on the contrary the system would have to simultaneously locate multiple tags in unchanged geometrical conditions As already mentioned a series of well determined points within the area in question was defined By using an accurate laser rangefinder and referring
41. operations described above private void CreateTableInDBFromAllTagsByIdTag string idTag string connectionString string ctStr SELECT INTO dbo idTag FROM UwbTagDB dbo data WHERE idTag idTag SqlConnection conn new SqlConnection connectionString SqlCommand command conn CreateCommand command CommandText ctStr conn Open command ExecuteNonQuery conn Close The CreateTablemDBFromAllTagsByIdTag method takes as input string containing the ID of the tag from which you want to select the in formation and performs a SQL SELECT query This query selects the rows in the date table containing all the data acquired by the RTLS and inserts them into a table called idTag if doesn t exist is it created filtering for the ID value of the tag passed as input private void CreateTableFromSelectedTagByTimePeriod string idTag double hours string 102 CHAPTER 6 THE CASE STUDY connectionString string startTime DateTime Now AddHours hours ToString string ctStr SELECT INTO dbo Last hours hours FROM UwbTagDB dbo idTag WHERE Date BETWEEN startTime AND 4 DateTime Now SqlConnection conn new SqlConnection connectionString SqlCommand command conn CreateCommand command CommandText ctStr conn Open co
42. out by the object Sensor which once processed the received message by the Tag object sends the trackedTag t1 ti 1 i for t 1 N and N number of sensors message to the Location Platform object This message indicates the transmission of the angles of arrival i and time of arrival ti received by the tags Also the activation relative to the object Sensor at this point is interrupted since the task of the sensors in the localisation process ends All the remaining operations are performed by the Location Platform ob ject without interacting with other objects in fact the next sent messages are all recursive These transactions consist in i the computation of the two positioning techniques used by the RTLS computeTDOA t1 ti and computeAOA 1 i ii computation of the tag position in the three coordinates x y and z computeTagLocation TDOA AOA and iii updating and storing of the new data acquired UpdateAndStore Data TagLocation As soon as the last step of the localisation process is completed both the activation of the Location Platform object and the lifeline of the entire process will be interrupted and the diagram is complete 6 2 3 RILS installation After the study of the system requirements and functionality it was possible to proceed with the installation phase of the RTLS As described in section 5 3 3 the installation process is a very impor tant step in which a special care should be
43. provided in order to avoid to compromise the system performances In the following sections the description of each step of this process is reported 6 2 3 1 Site survey The initial step of the installation process consisted in carrying out an in spection at the candidate location to be the object of the study in order to assess the environmental and ambient features where it was decided to implement the localisation system Figure 6 8 shows a picture taken during this phase The presence of metal lic structures walls and slurry accumulation involved the preparation of an 62 CHAPTER 6 THE CASE STUDY appropriate protective equipment from any accidents during the assembly phase but however it did not result in threats against the performances of the system due to the eventual reflection scattering and attenuation degree in presence of electromagnetic waves Figure 6 8 Photo of the study area In this initial phase a further study on the environmental features was carried out within the area in question datalogger connected to tempera ture and humidity sensors anemometers and globe thermometers shown in Figure 6 9 was installed in order to determine the suitability of the location to be the object of the experiment The comparison of the measured values over a period of about a month with those declared in the hardware features tables of the system section 5 3 1 1 confirmed the feasibility of the test in this environme
44. represent a starting point from which take inspiration for a study aiming at defining rules capable of generalize the potentiality offered by the software in multiple contexts such as in addi tion to estrus detection lameness detection effects of cows mobility on milk production etc References Agosta G 2012 De mizione di una metodologia di analisi del comporta mento di bovine da latte allevate a stabulazione libera in relazione al microclima Unpublished doctoral dissertation DiGeSA University of Catania Catania p 141 lvarez C amp Cintas C 2010 Accuracy evaluation of probabilistic loca tion methods in uwb r d Unpublished master s thesis Department of Electronic Systems Aalborg University Aalborg lvarez C N Cin tas C C 2010 Accuracy evaluation of probabilistic location methods in UWB RFID Department of Electronic Systems Aalborg University Aalborg p 111 Barbari M Conti L amp Simonini S 2008 Spatial identification of animals in different breeding systems to monitor behavior In International livestock environment symposium Vol VIII p 1 6 Iguassu Falls Brazil 31 August 4 September 2008 Bava L Tamburini A Penati C Riva E Mattachini G Provolo G et al 2012 Effects of feeding frequency and environmental conditions on dry matter intake milk yield and behaviour of dairy cows milked in conventional or automatic milking systems talian Journal of Animal
45. screenshot of the UI of the tool in the running state On the left the tag IDs which are currently present within the monitored area are listed and the total number of them is shown immediately above Selecting one of the IDs in the list the corresponding values of i Current Tag ID ii Update Time iii Current X iv Current Y v Current Z vi Standard Error are shown in the near textbox On the right the section Map which is a reproduction of the plan view of the study area is shown The coverage area of the sensors the four red circles is bounded by thick red lines The position of the selected tag is represented by a graphic element blue filled circle above which the three last digits of the ID are shown WS University of Catania Department of Agri food and Environmental Systems Management ILE 020000096004 Nn 020000121254 120000086023 E Update 93 99 2013 19 19 10 624 020000121239 time 020000139187 eda Current x 9 374 020000140053 Current 2 2 01019954681396 Figure 7 1 UI of the tool for RTLS data storage In Figures 7 2 and 7 3 instead two screenshots of the UI of this tool in two particular cases are shown 7 1 RTLS PERFORMANCES 125 Figure 7 2 in particular shows the data acquired from the RTLS for one of the tags applied to the cows in the case where the cow in q
46. section the operating specifications i e the principle of function ing of the Ubisense system will be explained In other words you will see how Ubisense components interact during the localisation process according to the theoretical principles detailed above Figure 5 14 shows the way in which communication occurs between sen sors and tags and an idea of how sensors use combined positiong technologies to calculate the location of the tag in a precise manner Master Sensor AOA master TOA master Slave Timing Synchronization via Cable Sensor TOA siave Figure 5 14 Principle of functioning of the Ubisense system The sensors are asking the tags to generate a UWB pulse through conventional radio signal RF At the time to the tag transmits the pulse and once received by the sensors calculates the Angles Of Arrival AOAs and o Sensors are synchronized via cable then Time Difference Of Arrival TDOA between each pair of sensors is calculated From the combination of TDOA and AOA position of the tag is calculated Heathcote 2011 We can summarize the process of estimating the location of a tag through the following steps 1 Sensors within the cellular Ubisense network communicate with the tags on conventional Radio Frequency RF signals green arrow in Figure 5 14 and ask to generate a UWB pulse depending on their profile up to 20 times a second 2 At the time to tags answe
47. system with f smaller than 2 5 GHz must have a fractional bandwidth larger than 0 2 Figure 5 2 Sahinoglu et al 2008 Power Spectral Density Frequency Figure 5 2 Characteristics of a UWB signal absolute bandwidth B is at least 500 MHz or fractional bandwidth byrac B f is larger than 0 2 Gezici et al 2005 5 1 ULTRA WIDEBAND SYSTEMS 23 5 1 1 2 Shape of the impulse The duration of the waveform in a UWB system is typically very short usually on the order of a nanosecond from 0 25 ns to 1 25 ns due to its large bandwidth These ultra short pulses have a low duty cycle In other words the ratio between the pulse transmission instant and the average time between two consecutive transmissions is usually kept small The typical shape of the pulse used in the UWB technology is that of the double Gaussian gaussian doublet Some common UWB pulse shapes include derivatives of the Gaussian pulse wavelet pulses Raised Cosine pulse and pulses based on modified Hermite polynomials For example the second derivative of the Gaussian pulse is expressed as w t Meme 5 5 where gt 0 and are parameters that determine the energy and the width of the pulse respectively In Figure 5 3 a unit energy pulse is plotted according to 5 5 0 4 ns and width of around 1 ns Sahinoglu et al 2008 104 Amplitude M RR Sea 0 5 04 03 0 2 0 1 0
48. systematically nearly the same zones of the study area in the same time intervals analysed in each of the days chosen for the study with the exception of the day where estrus occurred In fact during this day a distribution of points in a much wider area was observed In addition by observing the color of these points it was evident that the animal showed during the day of estrus a more sustained motor activity than that recorded in the other days The use of CowSpeedGraphs confirmed this observation as in the range where the cow was in estrus instantaneous velocities above the average of those obtained on other days were recorded 159 From these considerations it could be deduced that the behavior of the cow during the week analysed took on average was ripetitive except for the day when the state of estrus manifested With regard to cow estrus detection Rorie et al 2002 studied the ap plication of commercially available electronic estrus detection technologies to reproductive management of cattle and demonstrated that pedometers are the most applicable devices to lactating dairy cattle and have a higher ac curacy and efficiency when combined with visual observation Also Roelofs et al 2005 adopted a pedometer for estrous detection as predictor for time ovulation in dairy cattle and demonstrated that pedometers can accurately detect estrus though not in real time The collection of estrus data in real time is of considerable imp
49. tag in turn responds to the sensors that sent the request The Sensor class is also connected to the Location Platform class with multiplicity 1 because system can have one or more sensors while the Location Platform class has a relationship of multiplicity 1 against the Sensor class since a RTLS can be constituted by a single Location Platform This finally results to be a class consisting of more than one class with multiplicity 1 to 1 since there is only one Location Platform and every class which composes it has no other instances In particular the classes associated with the composition relation ship with Location Platform class are Location Engine Configuration Map Service Manager Security Manager Platform Control Service Installer and Site Manager 6 2 2 3 State diagram In UML language a state diagram is used to indicate the model type that shows the states of a system and the events that trigger a transition from one state to another This diagram shows how a system responds to external and internal events It is based on the assumptions that a system has a finite number of states and that events may cause a transition from one state to another Sommerville 2010 58 CHAPTER 6 THE CASE STUDY Graphically as shown in Figure 6 6 the states of a state diagram are represented by a rectangle with rounded corners The symbol that shows the transition from one to another state is drawn by using a continuous line that
50. the tag from this state it is necessary to simply move it or press the button This feature however was not used in our test because small movements or vibrations of the body of the cow even when it had not moved would activate the accelerometer present in the tag and then woke up the tag The LEC software tool provides the user with the management of the events recorded by the RTLS i e informations related to a particular tag being localized in a specific section of its User Interface UI as shown in Figure 6 18 Monitored sensor cells LocationEngineCell 0001 Event 309 gt Slot 396 Tag 020 000 000 008 Deka 16 36 87 43 71 1 10 eee a z Tag Maoa rTDOA Power PPO code C Leave trail Figure 6 18 LEC monitoring Controls The Monitored sensor cells list shows which cells are being monitored ANALYSIS AND INSTALLATION OF UBISENSE RTLS 73 The Tag text box allows you to filter the tags shown on the Cell Map The Recent tags button when clicked shows a history list of the most recent tag IDs you have entered The slider and associated controls allow you to rewind recorded events Every location acquisition by the tag generates an event in this list The Clear button clears all recorded events The text boxes show information about the selected event The slot number refers to the timeslot Delta means the difference from the previous timeslot for an event for that tag Th
51. video recording system 6 4 LOCALISATION AND IDENTIFICATION PERFORMANCES OF UBISENSE UWB RTLS 6 4 1 Automatic data preprocessing The construction of panoramic plan view images of the area under study made it possible to verify the planimetric position of each tag provided by the RTLS The data analysed in the following section were obtained from the video recordings carried out on 2nd August 2011 during a time interval of about 54 minutes Within this time interval two different cow behaviours were ob served The first time interval between 06h 26m 49s and 06h 53m 39s about 27 minutes included cow localisation during the feeding activity at the manger whereas the second time interval between 11h 35m 37s and 12h 02m 27s about 27 minutes regarded cow lying behaviour in the stalls According to one of the objectives of this thesis work which consisted in the evaluation of RTLS accuracy to monitor the feeding and lying activities of the cows the service alley adjacent to the second row of stalls was not framed by the multi camera system Therefore if the cow was in the service alley adjacent to the second row of bunks and therefore was not caught on the camera the operator had the ability to report such an event and the 88 CHAPTER 6 THE CASE STUDY corresponding measurements were excluded from subsequent analyses Firstly two different datasets were defined one contained the tag posi tions measured by the RTLS section 6 4 1 1
52. was rejected as random noise In Figure 6 15 power thresholds red horizontal lines are shown in relation to each sensor and noise signals are displayed green waves This operation made it possible to distinguish in the presence of a tag the signal noise from a UWB signal associated with a tag The last step before the calibration of the system consisted of the virtual reconstruction of the map of the sensors inside the LEC Figure 6 16 A reference system was determined within the graphical section of this software 68 CHAPTER 6 THE CASE STUDY File Edit View Log Sensor and Cells Sensor Status Filters Tags Owners Log 4b LL Dres Message m amp 06 16 11 11 47 55 unknown bootstrap 00 11 CE 00 39 F5 192 168 0 100 2823 Jauto revc master elf block 200 06 16 11 11 47 55 LocationEngineCell 0001 sensor timing cable error report for previous 15 seconds A DRE 0001 sensor proportion of slots with the wrong timeslot length 1000000ppm LocationEngineCell LocationEngineCell 0001 sensor proportion of slots without good set of phase samples 1000000ppm LocationEngineCell 0001 sensor proportion of slots without stable timing circuit phase 1000000ppm bootstrap 00 11 CE 00 39 F5 192 168 0 100 2823 auto_revc_master elf block 300 06 16 11 11 47 58 unknown 06 16 11 11 47 59 LocationEngineCell 0001 sensor timing cable error report for previous 15 secon
53. which have above them the last three digits of the respective tags are coloured according to the specific tag colour Each label which is used to determine the true position of a tag Figure 7 8 is characterized by the ID of the tag to which it is associated and from the colour chosen for the text according to the tag colour The coordinates of the cow true position displayed on the top right corner of the UI are highlighted according to the tag colour In Figure 7 9 the computation of the true position of tag with ID 004 was exemplified The association of a label to each cow make it possible for the operator to place the label 04 on the cow s head by a drag and drop operation The cow true positions shown in the top right section of the UI were automatically stored after clicking on the Save button 7 1 RTLS PERFORMANCES 133 gt University of Catania Department of Agri food and Environmental Systems Management Day Select 2011 08 01 Control Goto Curent Postion 8586 26827 lt lt 30 ts s20 TAG 08 Q Track Bar Postion 8586 image s Date 1 8 2011 10 48 3 D prr Eg 1D 251438 Tags Data identikit Y z WONG Y z 3553 10439 1818 893 4107 075 8 0000157312359078787 8 0000261390639934689 Date 01 08 2011 10 48 03 8 E s TAG 254 9 3 MM 9305 10863 0993 MM 55 102 0 93 5 St
54. x Y z x Y z 3633 11667 0631 iM 4593 7851 0742 Std tur 0 00328421732410789 0 00219549890607595 Date 23 08 2011 10 31 02 23 08 2011 10 31 02 TAG 53 TAG 51 X Y Z X Y z 365 6295 108 Std Ber 000067040883004665 29 Date 23 08 2011 10 31 02 Date Figure 7 6 Software UI for the computation of the true positions of UWB tags At the top left corner of the UI under the Day Select tab there is the Control tab which consists of Go to textbox suitable to enter from the keyboard the number of records relative to the loaded input file In this textbox it is possible to write the position required for the visual recognition of the position of each tag The Current Position is displayed below Trackbar to move to the desired record to be loaded by moving the cursor along the horizontal line directly with the mouse The current rackbar Position is displayed below Four buttons to move 30 record backwards 1 record backwards 1 records forward or 30 records forward from the current record respec tively The software automatically checks with respect to the position of the current record when enabling or disabling these buttons appro priately If for example the current record is the last record of the selected day the last two buttons are automatically disabled because since no other records to follow any pressing of one of t
55. 0 1 0 2 0 3 0 4 05 Time ns Figure 5 3 The UWB pulse gaussian doublet according to Equation 5 5 with 0 4 ns and width of around 1 ns Sahinoglu et al 2008 5 1 1 3 Spectral analysis communication can not be established by sending a single pulse since it alone is not capable of transmitting a quantity of information Therefore 24 CHAPTER 5 UWB RTLS a train of pulses that appear randomly in time to optimize the spectrum usage represents a transmission that carries information Figure 5 4 Mei 2003 Due to their very large portion in the frequency spectrum UWB signals are likely to cause interference with other systems As seen in Figure 5 4 the pulse train frequency takes the form of a noise signal noise like which minimizes the effect of the interference of the UWB signal if there are other signals that have been transmitted within the bandwidth occupied by UWB system Di Noia 2010 For example frequency allocation of some wireless systems is shown in Figure 5 5 Sahinoglu et al 2008 Random Postion Pulse Train Spectrum of Random Postion Pulse Train 100 150 200 3 gm i B a m E gom 2 an 400 B 1 A cU 65 7 74 a B5 3 95 10 Time 2107 ts Frequency 10910 Figure 5 4 Random positioned pulse train and its frequency spectrum noise like Mei 2003 Emitted signal power oe RERO Frequency GHz 1 6 1 9 2 4 3 1 5 8 10 6 Figure 5 5 Fre
56. 0706759048625827 21 07 2011 12 59 5 564 020000240053 10 945 8 932 1 1917 0 0003323215248076415 21 07 2011 12 59 55 780 020000140051 10 574 5 98 0 992 0 00118449644651264 21 07 2011 12 59 56 429 020000140053 10 746 187 1 45 0 00146156852133522 21 07 2011 12 59 6 64 020000140051710 2975 928 0 795 0 001822530 0332279 21 07 2011 12 59 57 294 020000140053 10 491 8 837 1 468 0 000297927384963259 21 97 2011 12 59 57 510 020000140051 10 513 5 979 0 818 0 00120969442650676 21 07 2011 12 59 58 159 020000140053 10 709 9 206 1 58470 000800565117970109 21 07 2011 12 59 58 375 020000140051 10 549 5 832 0 707 0 002771755269780755 21 07 2011 12 59 59 23 020002140053 10 83 9 066 1 46 0 00018 345852514729 21 07 2011 12 59 59 240 020000140051 10 524 5 971 0 857 0 000663391198031604 21 07 2011 q 888 2020000240053 10 75358 71871 45570 99029S652396744117 953 02000023 40051710 48275 777 5 248505689 2 2005 a Figure 6 20 Format of the output produced by the tool acquisition date acquisition time tag ID x y z standard error 6 2 5 1 Populating Database Simultaneously with write operations on the file aimed to disk storage of data acquired by the RTLS an extension of the software described above was specifically developed in this thesis work in order to automatically populate a database with these data and allow its subsequent processing Figure 6 21 shows the flow chart related to the operations implemented by the sof
57. 13 5 14 5 15 5 16 5 17 6 1 6 2 6 3 6 4 6 5 6 6 6 7 6 8 6 9 LIST OF FIGURES Add references in a Microsoft R Visual Studio Express project Step 1 first click with the right mouse button on the References of a project and then the left mouse button on Add Reference 41 Add references in a Microsoft R Visual Studio Express project Step 2 select the desired file with dll extension present in one of the folder to install the software Ubisense and click OK Links to the references in the project in order to be able to use the classes implemented in these libraries will be automatically created 42 Principle of functioning of the Ubisense system The sensors are ask ing the tags to generate a UWB pulse through conventional radio signal RF At the time to the tag transmits the pulse and once received by the sensors calculates the Angles Of Arrival AOAs and Sensors are synchronized via cable then Time Difference Of Arrival TDOA be tween each pair of sensors is calculated From the combination of TDOA and AOA position of the tag is calculated Heathcote 2011 43 Screenshots del Location Engine Configurator tool for the visualization of 3D e 2D maps respectively Green lines represent the AOA vectors from each sensor Blue hyperbolic curves represent the hyperbolic localisation Red dot is the tag position Heathcote 2011 44 Ins
58. 13 Development and evaluation on a rfid based traceability system for cattle beef quality safety in china Food Control 31 314 325 Finkenzeller K 2003 R d handbook Fundamentals and applications in contactless smart cards and identi cation John Wiley amp Sons Ltd The Atrium Southern Gate Chichester West Sussex PO19 850 England Firk R Stamer E Junge W amp Krieter J 2002 Automation of oestrus detection in dairy cows review Livestock Production Science 75 219 232 Gezici S Tian Z Giannakis B Kobayashi H Molisch F Poor H V et al 2005 Localization via ultra wideband radios IEEE Signal Processing Magazine Heathcote J 2011 Real time location systems to track people and assets Retrieved June 2013 from https connect innovateuk org c document_library get_file p_l id 3171260 amp folderId 4121020 amp name DLFE 34306 ppt Huhtala A Suhonen K M kel P Hakojarvi M amp Ahokas J 2007 Evaluation of instrumentation for cow position ing and tracking indoors Biosystems Engineering 96 3 399 405 Available from http www sciencedirect com science article pii S1537511006004053 Ilie Zudor E Kem ny Z Blommestein F van Monostori L amp Meulen A van der 2011 A survey of applications and requirements of unique identification systems and rfid techniques Computers in Industry 62 221 252 Ipema A Ven T van de amp Ho
59. 6 11 11 48 37 LocationEngineCell 0001 sensor 06 16 11 11 48 37 LocationEngineCell 0001 sensor 2003 2007 Ubisense Limited All Rights Reserved 06 16 11 11 48 37 LocationEngineCell 0001 sensor 06 16 11 11 48 37 LocationEngineCell 0001 sensor Build 2526 Dec 11 2009 20 41 21 06 16 11 11 48 37 LocationEngineCell 0001 Serial Number 1866 6951 146 4888 06 16 11 11 48 37 LocationEngineCell 0001 06 16 11 11 48 37 LocationEngineCell 0001 jer Configured For 37Hz update rate 06 16 11 11 48 39 LocationEngineCell 0001 sensor programming radio 06 16 11 11 48 39 unknown bootstrap Downloading auto standard 8 hex to 192 168 0 100 2310 3 06 16 11 11 48 39 0 1 unknown bootstrap 192 168 0 100 2310 initialising or missed packet 06 16 11 11 48 39 192 168 0 105 unknown bootstrap 00 11 CE 00 39 F5 192 168 0 100 2310 Jauto_standard_8 hex block 1 lt tareas Figure 6 13 Log window of the LEC Sensor and Cells Sensor Status Fiters Tags Owners Log f Aem 0 5 108 t _lastSeen IPAddress Flags 00 11 00 38 87 LocationEngineCel 0001 06 16 11 12 01 55 192 168 0 102 00 11 CE 00 38 D6 LocationEngineCell 0001 06 16 11 12 01 57 192 168 0 103 00 11 CE 00 39 9F LocationEngineCell 0001 06 16 11 12 01 59 192 168 0 101 00 11 CE 00 39 F5 LocationEngineCell 0001 06 16 11 12 02 01 192 168 0 100 timing source master HE Figure 6 14 The Sensor Status shown in the LEC after sensors boot operation
60. 7 0200000860904 7 584 1 67 0 486 0 003997 0296 68275 020000086004 7 984 1 6 0 486 0 00399 60296568275 020000086004 035 1 399 2 001 0 0108372449012233 020090086094 2 035 1 349 2 001 0 0108372448012333 020000084004 3 035 1 349 0 001 0 0108372448012233 020000006004 6 484 1 264 1 803 0 0007748992351565 020000096004 6 404 1 25 Q 0007 40089633156 020000086204 17 577 3 a LI LI Ld 4 4 oe e oe Ub oU o o m mmm mao mmT n sT Figure 6 35 Output format after the matching operation 6 4 2 The dataset of the true positions of UWB tags A specific software which was developed by using Microsoft R Visual C Ex press framework NET allowed visualisation of tags within each panoramic image by using graphic elements points For each tag the coordinates of the related graphic element in a panoramic were expressed in number of pixels and obtained through a linear conversion of the real tag coordinates for i 1 n where n is the number of measurements obtained by the RTLS expressed in meters The visual recognition of the position of each tag was carried out by an operator via the software An identikit of each dairy cow of the herd was made available to the operator in a section of the software interface in order to facilitate this activity Once a cow had been identified the operator
61. 8773 0200001 Current Y 5 546 Current 2 0 345 Std Error 2 71595692634583 Hide Figure 7 3 UI of the software for storing data acquired by the RTLS The displayed tag is applied to a cow in Lying 7 1 3 The sequence of panoramic top view images and the building of the dataset of tags true positions To obtain the fully rectified panoramic top view images of the two selected functional units i e feeding area and resting area the installation of 10 cameras was required according to a previously defined methodology Porto et al 2013 In the chosen time intervals the verification of 10742 tag positions was carried out through the use of 1600 panoramic images acquired by the video recording system The dataset of tags true positions was obtained by using the specifically developed software By using this software the accuracy assessment took about 40 hours of operator s work In the next section the user interface of this software is described 7 1 RTLS PERFORMANCES 127 7 1 3 1 User Interface of the software for the computation of tags true positions dataset The software specifically developed for the computation of the dataset of the true positions of UWB tags has a user friendly User Interface T his software provides a quick accurate automatic and interactive tool designed to take advantage of the technologies used in this study for the localisation and the video re
62. 9 6 30 6 31 Plan a and section b of the area of the barn under study with sensors Ia OE sched de Goa doe A out Ried Re ee Sensors IP30 Serie 7000 used in the trial a front side of a sensor b back side of Master sensor with Timing Cables black and power cable white c sensor covered with protective plastic material Scheme of the RTLS hardware used in the trial Log window ofthe LEC orior ot Rs The Sensor Status shown in the LEC after sensors boot operation The table in the figure shows the MAC address of the sensors the cell to which they belong the current status date and time in which they were active before the last update the IP address and any flags respectively The power thresholds of each sensor indicated with a red horizontal line allow to distinguish the noise signal green wave from the UWB pulses transmitted by the tags 060 eo xXx BO 2D sensors map in the graphical section of the LEC Each sensor is iden tified by its MAC address represented by a string of 12 hexadecimal digits Battery status and update rate of each tag LEG monitoring Controls s s aua ea REE e Se Y vo Emgu CV Architecture Overview EmguCV 20013 Format of the output produced by the tool acquisition date acquisition time tap ID se yea standard error 2 2 64 04 se Yee es Flow chart of the software for the a
63. EED 0 Cow s Speed m s 0200 0400 0600 10 00 1240 1440 Time m Tag Id 020 Date n Tag Id 020 Date 16 08 2013 In 16 08 2013 Interval CLTime terval CLTime Figure 7 18 CowHeatMaps and CowSpeedGraphs relative to the tag with ID 020 in CLTime interval distinct by day 3 3 study area by the cow in question were mainly those corresponding to the cubicles in the middle or in the left in particular and in smaller quantities at the manger referring always to the Figure 6 10 On the day of 12th August 2013 however the distribution of points in the space mainly involved as well as at the manger the service alley the left and right passage lane and the feeding alley as already observed in the CF Time interval Furthermore also in this case by observing the coloration of these points it was evident that the animal to which the tag with ID 020 was associated on 12th August 2013 showed a motor activity more sustained than that recorded in the other days Furthermore in Figure 7 18 the CowSpeedGraphs associated to the CowHeatMap are shown By observing these graphs also in this case especially in the time interval ranging from 11 00 am and 11 30 am when the CLI usually shows maximum values on 12th August 2013 higher values of instantaneous speed were recorded compared to those obtained in the other days Finally figure 7 19 shows the CowSpeedGraphs relative to the tag with ID 020 distinct by day
64. ER 6 THE CASE STUDY Importing Ubisense libraries a ROG N using Ubisense UBase using Naming Ubisense UName Naming using Ubisense ULocation using Ubisense ULocation CellData using System Media Creation of naming schema object and filewriter output create naming schema object private static Ubisense UName Naming Schema naming schema null filewriter output TextWriter log new StreamWriter date csv true Initialization of naming schema object and connection as a client naming schema new Ubisense UName Naming Schema false naming schema ConnectAsClient Creation of multicell object and loading of all Ubisense cells a a R O MultiCell multicell new MultiCell SortedDictionary lt string Cell cells multicell GetAvailableCells foreach Cell cell in cells Values multicell LoadCell cell true Getting all objects out of Ubisense DB using ReadTransaction xact multicell Schema ReadTransaction foreach Location RowType r in Ubisense ULocation CellData Location object xact LocationOutput r Definition of update event handler used if an Ubisense object moves Ubisense ULocation CellData Location AddUpdateHandler multicell Schema CellData Update Event Handler of Ubisense database location update private void CellData Update Location RowTyp
65. F5 192 168 0 100 2823 Jauto revc master elf block 1000 06 16 11 11 48 22 192 168 0 105 unknown 39 F5 192 168 0 100 2823 auto revc master elf block 1100 06 16 11 11 48 25 192 168 0 105 unknown bootstrap 00 11 CE 00 39 F5 192 168 0 100 2823 revc master elf block 1200 06 16 11 11 48 28 LocationEngineCell 0001 sensor timing cable error report for previous 15 seconds 06 16 11 11 48 28 LocationEngineCell 0001 sensor proportion of slots with the wrong timeslot length 1000000 06 16 11 11 48 28 10001 sensor proportion of slots without a good set of phase samples 1000000 06 16 11 11 48 28 LocationEngineCell 0001 sensor proportion of slots without stable timing circuit phase 1 06 16 11 11 48 28 unknown bootstrap 00 11 CE 00 39 F5 192 168 0 100 2823 auto revc master elf block 1300 unknown bootstrap 00 11 CE 00 39 F5 192 168 0 100 2823 auto revc master elf block 1400 LocationEngineCell 0001 sensor timing cable error report for previous 15 seconds LocationEngineCell 0001 sensor proportion of slots with the wrong timeslot length 1000000ppm LocationEngineCell 0001 sensor proportion of slots without a good set of phase samples 1000000 LocationEngineCell 0001 sensor proportion of slots without stable timing circuit phase 11 06 16 11 11 48 37 LocationEngineCell 0001 sensor Ubisense Location Engine Sensor Runtime Version 2 1 192 168 0 100 0 00 11 00 39 5 06 1
66. From animal monitoring to early warning systems through the internet of things In Precision livestock farming 2013 leuven belgium Schwartzkopf Genswein K Huisma C amp McAllister T 1999 Validation of a radio frequency identification system for monitoring the feeding patterns of feedlot cattle Livestock Production Science 60 27 31 Shahi A Aryan A West J S Haas C T amp Haas R C 2012 Deterioration of uwb positioning during construction Automation in Construction 24 0 72 80 Available from http www sciencedirect com science article pii 0926580512000246 Sommerville I 2010 Software engineering 9th ed Harlow England Addison Wesley Sowell B Bowman J Branine M amp Hubbert M 1998 Radio frequency References 169 technology to measure feeding behavior and health of feedlot steers Applied Animal Behaviour Science 59 277 284 Steggles P amp Gschwind S 2005 The ubisense smart space platform A Ubisense White Paper Stephan P Heck I Kraus P amp Frey G 2009 Evaluation of indoor positioning technologies under industrial application conditions in the smartfactorykl based on en iso 9283 In Proceedings of the 13th ifac symposium on information control problems in manufacturing p 874 879 Togersen F Skjoth F Munksgaard L amp Hojsgaard S 2010 Wireless indoor tracking network based on kalman filters with an application to monitoring dai
67. Hz the maximum resolution time of the pulse is of the order of 133 picoseconds Therefore when a pulse is detected it is possible to know within 133 picoseconds the time of flight impulse with an uncertainty of 4 cm For systems with a bandwidth of 500 MHz however the corresponding time resolution is 2 nanoseconds which corresponds to an uncertainty space of about 60 cm With any of the UWB signal therefore it is potentially possible to obtain an accuracy less than one meter in the localisation Oppermann et al 2005 There are several techniques to determine the position starting from es timates of the time of arrival time of flight or angle of arrival In the next section some of them will be analysed 5 2 3 1 Positioning techniques for UWB systems Most of the positioning schemes estimate the position of objects to locate es tablishing geometric relationships between the transmitters and the receivers This is achieved through special techniques that are based on timing infor mation Time of Arrival ToA and Time Difference of Arrival TDOA Received Signal Strength RSS and Angle of Arrival AoA There also exist approaches that utilize location dependent characteristics instead of geometric relations of these measurements to find the position Such tech niques are usually called location fingerprinting Yan 2010 5 2 LOCALISATION IN ULTRA WIDEBAND SYSTEMS 29 Some techniques for UWB positioning systems are analysed
68. Integration of devices and sensors can be achieved by two methods for more straightforward tasks many parts of the platform are accessible via Application Programming Interface API presented in C or COM for more complex tasks where tighter integration is required it is possible di rectly to use the Ubisense data modelling language used to build the rest of the Ubisense system Steggles amp Gschwind 2005 5 3 1 2 Middleware software architecture Middleware software architecture is constituted by the Ubisense Location Platform a suite of software components that enable location sensors and tags to be set up calibrated and configured into cells and objects using a simple Graphical User Interface GUI Ubisense tag locations are sent via standard Ethernet cable or wireless LAN to the Location Engine which processes the data and passes the infor mation via an industry standard API to applications The Location Platform scales seamlessly from a laptop to a multi CPU cluster Exactly the same software can be run on a single machine for soft ware development or demos or deployed across a cluster to support huge installations generating tens of thousands of location events per second The Location Platform has a service oriented architecture All services are crash tolerant restart tolerant and cluster friendly and have been proven through years of testing in large scale real time control applications in as sembly plants transit depot
69. NSE RTLS 85 Location Engine Config agg File Edt View Cel Sensor Available Events Help ers tog T Figure 6 28 3D view of the study area The displayed tag is one of the tags applied to the cows Sensor and Cells Sensor Status nter Cel Plan d LocationEngineCell 0001 C Available Sensors Monitored sensor cells LocationEngineCell 0001 Tag 020 000 086 020 Recent tags Recorded events v Event 9999 6 Clear Slot 45 025 194 020 000 086 020 Delta XYZ 9 89 11 03 1 02 Time 12 30 42 Celi 0001 Bta Omoa rower pro uws code L teave trail 7 62825 9 26121 Location Engine Config Fie Edt Map Cell Sensor Available Events Help ars tog Tm Figure 6 29 3D view of the study area The displayed tag is the reference tag Sensor and Cells sensor Status Alter Cel Pian C Available Sensors Monitored sensor cells LocationEngineCell 0001 Tag 020 000 086 008 Recent tags Recorded events J Event 9999 o 9e Slot 45 067 006 020 000 086 008 Delta XYZ 3 38 10 21 1 47 Time 12 49 32 Cell 0001 Baoa CPower Cipro uws code trail 1 74537 7 36974 86 CHAPTER 6 THE CASE STUDY Location Fngine Config Edit View Cell Sensor Available Events Help Sensor and Cells Sensor Status Fiter Cet Plan dll LocationEngineCel 0001 C Available Sensors Monitored sensor
70. PERFORMANCES 93 The tag positions acquired at the time t not included in the interval lt 3s t 38 were discarded The problem of matching these datasets was addressed by developing a special software in the C programming language This application by using the iterative binary search algorithm Demetrescu et al 2004 com puted the matching between the records of the two datasets to intersect distinguishing the three cases described above The input of this application were two folders one containing the text file generated during the Reor ganisation of the dataset phase described in 6 4 1 1 section and one that contains the text file generated in Exporting data from database phase de scribed in 6 4 1 2 section This software tool returned in output a folder for each tag containing as many text file as were the days in which the match ing of records belonging to the dataset of tags and records belonging to the dataset of images occurred Below is the code of the implementation of the iterative binary search algorithm used in the specifically developed tool N e O ON O OC RB WwW t public bool binarySearch array lt FileStructType gt v int dimvect _ int64 sought int i m found false jc 05 j dimvect 1 do m i j 2 if sought v m date found true index m else if sought v i date found true index i else if sought v j date found tr
71. Science e42 11 Bell N Miedema H Blajan I amp Mottram T 2013 Automatic measure ment of mobility score in dairy cows using walking speed Computers and Electronics in Agriculture in press Berckmans D 2013 Editorial In Precision livestock farming 2013 Leuven Belgium Brehme U Stollberg U Holz R amp Schleusener T 2008 Alt pedometer new sensor aided measurement system for improvement in oestrus detection Computers and Electronics in Agriculture 62 73 80 Crockford D 2008 Javascript the good parts First ed Yahoo Ed O Reilly Media Inc 1005 Gravenstein Highway North Sebastopol CA 95472 165 166 References Demetrescu C Finocchi I amp Italiano F 2004 Algoritmi e strutture dati The McGraw Hill Companies Di Noia M 2010 Algoritmi di localizzazione in sistemi ultra wideband Domain T 2013 Ultra wideband r d Available from http www timedomain com datasheets UWBInterference pdf Dygraph 2013 Opensource javascript dygraph library Retrieved 25 10 2013 from http dygraphs com EmguCV 2013 Emgu cv Opencv in net c vb c and more Retrieved September 2013 from http www emgu com wiki index php Main Page FCC 2002 Before the federal communications commission washington d c 20554 Available from http transition fcc gov Bureaus Engineering Technology Orders 2002 fcc02048 pdf Feng F Z J Wang Z Xu M amp Zhang X 20
72. TER 6 THE CASE STUDY than the threshold The choice of the value of 0 50 m for the threshold is founded on the following consideration Since the distance between the extremities of cow s ears is equal to about 0 60 m on average and since the tags were applied some to the left ear and others to the right ear the center point of the cow head was chosen as the origin and 0 30 m from the origin as threshold in each direction x and y which corresponds to about 0 42 m in two directions Finally it was decided to further increase the margin of error by choosing a threshold value equal to 0 50 m In Figure 6 36 the area of the green square with sides of 0 60 m represents the considered surface for the computation of the threshold adopted in the Metric C Figure 6 36 Computation of the threshold established for localisation error in Metric C 6 5 SOFTWARE TOOL FOR THE VISUAL ANALYSIS 99 Furthermore another area was determined red square in Figure 6 36 in order to perform the same analysis of the performances in the case when the technique of application of the tag using the collars subject of a study in progress was adopted In this case the average width of the head about 0 35 m was considered as the starting point for the computation of the relative threshold 65 AN AUTOMATIC AND REAL TIME SOFT WARE TOOL FOR THE VISUAL ANAL YSIS OF THE COWS LOCATION DATA IN FREE STALL BARNS A specific software which was developed by usi
73. TLS in the static test Table 7 1 and these obtained for each one of the eight tags analysed and for the reference tag ID 187 before outlier data cleaning Table 7 2 and after outlier data cleaning Table 7 3 were shown The performance of the RTLS resulted higher in static conditions than in the case of moving objects Tables 7 2 and 7 3 shows that the localisation error made by the RTLS for the reference tag is definitively lower than the errors computed for the moving tags An analogous observation was gathered in a test carried out in a railway tunnel Mok et al 2010 The static test carried out in this study section 6 2 6 confirmed this concept In fact the average value standard deviation of the means relative to x y and z coordinates resulted about 0 13m 0 14 m 0 11 m 0 11 m and 0 16 m 0 14 m respectively Table 7 1 shows mean x mean y and mean z obtained in this test for each point analysed Tables 7 2 and 7 3 show the minimum value the mean value the maxi mum value and the value at the 90 percentile of tag planimetric position errors obtained before and after outlier data cleaning respectively Before applying outlier data cleaning technique to the dataset the average values of the mean error and the error at the 90 percentile of the eight tags resulted about 0 56 m and 1 03 m respectively Outlier data cleaning allowed reduction of the localisation errors made by the RTLS and made data distribution more h
74. TLS performance is generally not dependent on cow behaviour as it was observed for other systems Porto et al 2012 Ipema et al 2013 The analysis of the behaviour of the two cows that were associated with the highest localisation errors tag ID 023 and tag ID 026 showed that especially in the second time interval i e between 11h 35m 37s and 12h 02m 27s the areas occupied by these two cows were almost equal in size and location and close to the boundaries Therefore a possible explanation of the higher errors obtained may be related to the low AoAs in that area that could cause lower performances of the RTLS However as regards the software tool specifically developed for the visual analysis of the cows location data in free stall barns promising results were obtained with regard to both the performance of the implemented features and its use for the purpose of estrus automatic identification The test in fact showed excellent performance in terms of Portability it was able to work in different environments it worked in all major browsers including IE8 This fundamental aspect allows you to have economic advantages as it can amortize the costs by transport ing the application in different environments Being a web application this software provides a guarantee regarding this feature Real time correctness it ensured at the same time a satisfactory logical consistency of the operations carried out and an acceptable
75. after having isolated tags with adhesive tape b d d Figure 6 24 Preparation of the collars a a tag installed in a collar using plastic cable ties b the outer side of the collar c the inner side of the collar d the eight collars after having isolated the tags with an adhesive tape 82 CHAPTER 6 THE CASE STUDY In Figure 6 25 instead a photo of the installation process of a collar equipped with a Ubisense tag to a cow is shown Figure 6 25 Application of a collar equipped with tag to a cow 6 2 8 Filtering algorithms Before proceeding with the storage of data acquired by the RTLS related to the localisation of the cows tags a final setting operation was performed by the software in the LEC tool The Location Engine supports different algorithms for estimating tag po sitions from sensor measurements Each algorithm can have a number of pa rameters that control the behaviour of the algorithm and can include known constraints on the motion of tracked tags The Location Engine allows a set of parameters to be stored as a filter that can be applied to individual tags or a range of tags Filters are defined in the Filters tab of Location Engine Config Ubisense 2008 2010 In other words you can customize the system according to the context in which it s installed in order to optimize resources and to improve per ANALYSIS AND INSTALLATION OF UBISENSE RTLS 83 formance It allows you to set a
76. ams are used when developing an object oriented system model to show the classes in a system and the associations between these classes A class is a category or group of objects that have similar attributes and similar behaviors Graphically as shown in Figure 6 5 a class is represented by a rectangle The class name by convention is a word with an initial capital letter and appears near the top of the rectangle An attribute is a property of a class It describes a set of values that the property can have when the objects of that class are instantiated A class can have zero or more attributes The list of attributes of a class is graphically separated from the name of the class to which it belongs by a horizontal line The attribute name can be followed by two points and the data type with which you want to represent e g string float int bool etc In Figure 6 5 the Sensor class for example has the attribute MacAddress String Each sensor in fact is characterized by a unique string that represents its MAC address A method is an action that objects of a certain class can accomplish Like the name of the attributes the name of a method is written with lowercase characters If the method name consists of several words then those words 56 CHAPTER 6 THE CASE STUDY Figure 6 5 Class diagram ANALYSIS AND INSTALLATION OF UBISENSE RTLS 57 are joined together and each of them except the first is written with the first
77. ange the update rate in real time depending on the position send feedback to users via tag s LEDs query tag battery status informa tion and button press action Two way communication is also used for presence detection control and telemetry Ease of maintenance The sensors are managed remotely via TCP IP and Ethernet standards for communication and configuration The sen sor firmware is easily upgradeable via the network to allow the instal lation of software release updated Tags Ubisense tags Figure 5 11 Table are small and robust devices that when applied to assets or people enable their real time location in 3D with an accuracy of 15 centimeters In addition to the localisation features tags include additional features such as a LED for easy identification a buzzer to provide messaging features a motion detector to instantly activate the 36 CHAPTER 5 UWB RTLS LED visible Strapping through enclosure slots y Button Mounting holes Figure 5 11 Ubisense Series 7000 Compact tag tags and buttons that can be associated with specific events defined by the software Tags technical specification Dimensions 38 cm x 39 cm x 16 cm Weight 25 g Temperature range Humidity range Update rate Peripherals Certifications Power supply Mounting options from 20 C to 60 C from 0 to 95 non condensing from 0 00225 Hz to 33 75 Hz can be varied dynamically under software control
78. as it comes to small files and in terms of compatibility with the majority of client systems The following source code shows the implementation of the operations performed by the software to create the html file m 2 QC private void WriteHtmlWithVelocityData string idCow string fileName C cowSpeedGraph_ idCow html StringBuilder content new StringBuilder content AppendLine html content AppendLine head content AppendLine lt script type text javascript src dygraph combined js gt lt script gt content lt head gt content AppendLine lt body gt content AppendLine lt div id graphdiv style width 800px height 500px gt lt div gt content AppendLine lt script type text javascript gt content AppendLine g new Dygraph content AppendLine document getElementById graphdiv content AppendLine ad string date string splitDate string splitDatel string time string day string month string year for int i 0 i lt instantSpeed2D Count 1 i date instantSpeed2D i Date ToString Replace splitDate date Split time splitDate 1 splitDatel splitDate 0 Split 116 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 CHAPTER 6 THE CASE STUDY da
79. ased on the implementation of two instruments the CowHeatMap and the CowSpeedGraph The sections 7 2 1 1 and 7 2 1 2 describe the main features offered by these tools respectively 142 CHAPTER 7 RESULTS 7 21 1 CowHeatMap CowHeatMap contains an archive suitable to colorize the heatmap based on relative data It means that if you are adding only a single datapoint to the dataset it will be displayed as the hottest red spot then adding another point with a higher count CowHeatMap will dynamically recalculate the color intensity At the configuration phase it is possible to specify the following properties in order to customize the heatmap instance radius optional Type Number It is the radius of a single datapoint in the CowHeatMap measured in pixels element required Type String HTML Element It either provide an element s id or the element itself which should contain the heatmap visible optional Type Boolean It specifies whether the CowHeatMap is visible or not gradient optional Type Object It is an object which contains colours ranging from 0 to 1 opacity optional Type Number 0 100 It provide the opacity of the CowHeatMap measured in percent 7 21 2 CowSpeedGraph In the CowSpeedGraph the chart is interactive In fact it is possible to execute the following actions on the chart Mouse over to highlight individual values Click and drag to zoom Double clicking will execute a zoom bac
80. avioural patterns A further improvement about the features offered by the software could regard to the implementation of functions for the computation of some quan tities useful to perform statistical operations such as the Maximum Speed Minimum Speed Average Speed Maximum Acceleration Minimum Accelera tion Average Acceleration Moving Duration Moving Distance Moving Av erage Speed etc 160 CHAPTER 8 DISCUSSION Part V CONCLUSIONS 161 Chapter 9 Conclusions The main objective of this thesis work which consisted in the evaluation of localisation and identification performances of a real time location system based on ultra wide band technology was achieved through a number of activ ities The system was analysed by performing a requirements analysis which includes functional and non functional requirements analysis and hardware and software requirements analysis The performances of the RTLS based on UWB technology considered in this research were assessed in operative conditions typical of the harsh environment of a free stall barn which was located in South Eastern Sicily Italy A site survey was carried out to study the environmental features of the barn under study and the sensor layout was designed on the basis of the recommendations provided by the producer of the system The system was then calibrated and subject to a static test to evaluate the performances in the specific breeding environment To perform the ass
81. btained after the zoom operation carried out in the previous figures 7 2 SOFTWARE TOOL FOR THE VISUAL ANALYSIS 145 Real time software tool for the analysis of the cow s location data 20000138187 ap UNIVERSITY DiGeSA lt OF CATANIA Figure 7 14 An example of CowHeatMap displayed in the visualisation section of the UL Real time software tool for the analysis of the cow s location data 20000138187 2013 08 15 23 53 58 COW SPEED 0 04 be WR w s Speed m s 236 BI 2340 46 148 2550 299 25 ap UNIVERSITY DiGeSA gt OF CATANIA Figure 7 15 An example of CowSpeedGraph displayed in the visualisation section of the UI 146 CHAPTER 7 RESULTS Real time software tool for the analisys of the cow s location data 20000138187 2013 08 15 23 51 42 COW SPEED 0 08 i E e Gh UNIVERSITY BOT OF CATANIA Figure 7 16 CowSpeedGraph obtained after the zoom operation carried out in Figure 7 19 7 2 3 Use case of the software tool The results obtained from the use case of the the software tool described in section are shown in Figures 7 17 7 18 and 7 19 Figure 7 17 in particular shows the CowHeatMaps and the CowSpeed Graphs relative to the tag with ID 020 in the CFTime interval distinct by day As already introduced in section 6 5 4 seven days ranging from 10th August 2013 to 16th August 2013 were chosen since in one of these
82. ccuracy provided by RFID systems based on HF and UHF technologies could be improved by using Ultra Wide Band UWB technology Alvarez amp Cintas 2010 The advantage of this technology is provided by the signal transmission mode which takes place by means of short duration pulses This mode of transmission provides UWB based systems with a low sensitivity to interference due to the absence of reflection of the wave itself In this chapter the features of an ultra wide band real time localisation system are described in detail 5 1 ULTRA WIDEBAND SYSTEMS A traditional UWB system is composed of two electronic components tags and readers which exchange information through radio waves and UWB pulses A tag is a device capable of transmitting a unique identification num ber when requested by a particular signal a reader is a device that by send ing radio waves urges the transmission of tag and receives the information transmitted The tags can be active passive semi passive or semi active A system using UWB technology for telecommunications and for locali sation has the following features Simplicity of implementation in a UWB system the modulation of the signal i e the set of techniques with which it is possible to adapt to the communication channel used an electrical or electromagnetic 19 20 CHAPTER 5 UWB RTLS signal to be transmitted without altering its information content takes place directly in the antennae Suc
83. ce in relation to the harsh environment analysed In fact no connection or communication failure occurred and few damages due to oxidation of PoE cables plugs were observed The protecting wrapping of the tags resulted suitable to the operative conditions of the trial since it avoided localisation data losses due to device failure Furthermore no infections occurred to cows ears due to tag applica tion However a further improvement was obtained by using the methodology of application with collars because it s a totally non invasive method offers the possibility to integrate additional devices for other studies guarantees a longer duration in time allow for easier interventions in case of maintenance of tags due to the simplicity of insertion removal of the collars The collars are also accessories easily commercially available and do not require high costs Since tag battery consumption is directly related to system configuration the RTLS acquisition time rate chosen in this study determined a tag battery duration of about two months 7 1 2 The software for storage of the data acquired by the RTLS During cow localisation phase in order to store the data acquired by the RTLS the specifically developed software described in section 6 2 5 was 123 124 CHAPTER 7 RESULTS launched This software at the same time allowed the user to view real time infor mations acquired from the RTLS through an interface Figure 7 1 shows a
84. considered reference system were written in the Location Engine Configuration tool then the system took readings of the AoA from the tag and yielded the orientation of the sensor from the average of the measurements set This calibration step was performed on each sensor in order to set their parameters of orientation pitch yaw and roll within the network The position of the tags used for this operation is shown in Figures 6 22 A central point within the study area close to the barycentre of the rectangle having as vertices the points where the sensors were installed was chosen in order to orient the sensors as symmetrically as possible between them As the cows mainly held the most central areas of the test area perform a calibration using as reference point a point placed on the central area would have ensured the best performance of the system for the case in question The second phase of calibration also called Cable calibration was similar to orientation calibration except that a cable offset was between two sensors Therefore the calibration used two sensors the sensor getting its cable offset and the reference sensor which was ideally the master Cable offsets were just some arbitrary number indicating the offset from the master so the master had a value of zero For this reason this latter calibration phase unlike the previous one was only carried out on each of the slave sensors The calibration process described above was
85. cording At program startup it is necessary to enter the input data which consists of the folder that contains the text files identified by the day of recording related to the location systems and the video recording system when the matching occurred and also grouped by tags This input is in fact the output of the software which carried out the matching between the tags dataset and the images dataset described in section 6 4 1 3 The input loading operation is described in Figure 7 4 gt University of Catania Department of Agri food and Environmental Systems Management Day Select Figure 7 4 User Interface at program startup By clicking on the Open button a win dow that allows you to search for the folder containing the input of the program is opened After the input data are loaded the software make it possible to choose the day of recording of the two systems for which you want to perform the 128 CHAPTER 7 RESULTS visual recognition of the position of each tag In Figure 7 5 a screenshot of the phase described above is shown gt University of Catania Department of Agri food and Environmental Day Select B Tn n n 1 8 2011 6 0 2 242852 X 2 X Y 2 T 2 Y 2 Figure 7 5 Day Select section of the tool A select box allows you to select the day of
86. cs and statistics of location information Server side computer with optimum operating system and hardware permanent Internet connection to communicate with the client Maintain a database of all known APIs Maintain location informations of all clients and update this whenever possible Display this information graphically Send this informations to the clients A fully functional Graphical User Interface GUI 6 2 1 2 Non functional requirements Non functional requirements as the name suggests are requirements that are not directly concerned with the specific services delivered by the system to its users They may relate to system properties such as reliability response time store occupancy and the constraints on the system implementation such as the capabilities of I O devices or the data representations used in interfaces with other systems Sommerville 2010 In this case study the non functional requirements are ANALYSIS AND INSTALLATION OF UBISENSE RTLS 53 Provide a pleasing interface and offer a robust and reliable service Be flexible because the users requests change over time A permanently available connection between the client and server response time as low as possible Be portable for the user in order to allow the use of the same application on different devices without requiring performing an installation each time 6 2 1 3 Hardware and software requirements The system would have to requir
87. cular as expected in this case the most frequented zones in the 150 a 14 020 Date 10 08 2013 Interval CLTime 1 1 2 3 c Tag Id 020 Date 11 08 2013 Interval CLTime e Tag Id 020 Date 12 08 2013 Interval CLTime CHAPTER 7 RESULTS 1000 120 1400 b Tag Id 020 Date 10 08 2013 Interval CLTime os 2013408711 00 00 02 COW SPEED 0 0200 0400 0500 0800 10 00 00 1400 d Tag Id 020 Date 11 08 2013 In terval CLTime 0130812 00 00 55 COW SPEED 0 0220 0600 0800 1000 120 1400 Time f Tag Id 020 Date 12 08 2013 In terval CLTime Figure 7 18 CowHeatMaps and CowSpeedGraphs relative to the tag with ID 020 in CLTime interval distinct by day 1 3 7 2 SOFTWARE TOOL FOR THE VISUAL ANALYSIS 151 g Tag Id 020 Date h Tag Id 020 Date 13 08 2013 13 08 2013 Interval CLTime Interval CLTime 20108 14 00 00 02 COW SPEED 0 Number of Location Events 02 00 0400 10 00 1200 1400 Time i Tag Id 020 Date j Tag Id 020 Date 14 08 2013 In 14 08 2013 Interval CLTime terval CLTime k Tag Id 020 Date 1 Tag Id 020 Date 15 08 2013 In 15 08 2013 Interval CLTime terval CLTime Figure 7 18 CowHeatMaps and CowSpeedGraphs relative to the tag with ID 020 in CLTime interval distinct by day 2 3 152 CHAPTER 7 RESULTS 2013 08 16 0000 03 COW SP
88. curacy is considered as an additional measure of accuracy for example 15 cm on the accuracy of 90 of cases The precision of a measurement system is the degree to which repeated measurements under unchanged conditions show the same re sults Two components of the precision are the reproducibility and the repeatability described below Reproducibility The quality of measurement that reflects the close ness of the results of measurements of the same quantity performed under different conditions i e in different laboratories at different locations and using different equipment is called the reproducibility of measurement Good reproducibility indicates that both the random and systematic errors are small Rabinovich 2010 Repeatability The quality of measurement that reflects the closeness of the results of measurements of the same quantity performed under the same conditions is called the repeatability of measurements Good 5 2 LOCALISATION IN ULTRA WIDEBAND SYSTEMS 27 repeatability indicates that the random errors are small Rabinovich 2010 Recall The recall also called sensitivity or true positive rate mea sures how often we find what we are looking for In other words is the proportion of actual correct measures which are correctly identified as such Cost In cases where a substantial number of elements must be equipped with devices for the localisation the cost is an important aspect A good remedy is the use of
89. d 020 Date 11 08 2013 Interval 24h 2013 08 13 00 00 02 COW SPEED 0 Cow s Speed m s a il lu il i BI Ji uL LL adal ei O MD 0600 1000 1200 1400 1600 1390 2000 220 Time d CowSpeedGraph Tag Id 020 Date 13 08 2013 Interval 24h 1 LL NN N i LT au TA T M my PR RT YI RARI POR O 0200 0600 1000 1200 1400 1600 1890 2290 Timc CowSpeedGraph Tag Id 020 Date 15 08 2013 Interval 24h 2013 08 16 00 00 03 COW SPEED 0 g CowSpeedGraph Tag Id 020 Date 16 08 2013 Interval 24h Figure 7 19 CowSpeedGraphs of to the tag with ID 020 distinct by day relative to all the 24 hours 153 154 CHAPTER 7 RESULTS Chapter 8 DISCUSSION The environment of the barn was characterised by the presence of slurry accumulation in the alleys as the cleaning was not automated but it was carried out once a day This condition facilitates production of gases that may enhance corrosion of materials On the other hand this effect is reduced by ventilation which is higher in barns with open sides In relation to the analysed environment the physical features of the Ubisense tags and the technique adopted to protect them in this trial allowed a good usability of the system This demonstrates that some problems mainly due to the dimensions and the weight of the RFID UHF tags which were highlighted in a pre
90. d Std E 8 000117613875772804 0 0020701892208308 Wek 8 i Date 01 08 2001104803 01 08 2011 10 48 01 a f E s Y z T X Y z Z 3 i Br V 4622 899 0268 z Std Std 2 bak 8 0 000135968060931191 amp Date Date 01 08 2011 10 48 03 TAG 239 AG 20 x Y x z le WM 3888 12975 128 8753 11581 0892 Sid Sd gt i 8 0 000581809203140438 29 0 000897705496754497 Date 01 08 2011 10 48 06 Dae 01 08 2011 10 48 03 TAG 53 V TAGS ZW s 302 054 Qd 00586813455447555 Bd E T Date 01 08 2011 10 48 03 pu CC Figure 7 9 The computation of the true position of tag with ID 004 The association of a label to each cow made it possible for the operator to place the label 04 on the cow s head by a drag and drop operation The cow true positions shown in the top right section of the UI were automatically stored after clicking on the Save button Another important feature was implemented in the software in order to provide greater precision to the operations of computation of the true position of tags This feature allows you to perform pan and zoom operations by using the mouse scroll wheel in the Map section as shown in Figure 7 10 Another example of the computation of the true position of tags is exem plified in Figure 7 11 The format and structure of the output of the software described above are the same as its input as described
91. dimensional data of each cow using colours was implemented as a functionality of this tool The different colour and colour intensity denoted the difference in sample density at a location Measurement of the instant speed of each cow over the time represented through an interactive graph was another functionality implemented in this software The results obtained by a use case of this software tool which was carried out in order to acquire useful informations related to the occurrence of estrus showed that a pattern related to the behaviour of the cow analyzed can be identified when the state of estrus occurs Moreover since it was designed to monitor cow behavior in real time it offers the ability by adding new control modules to notify any inconvenience through alert messages as a result of changes in dairy cow behaviour and therefore it is possible to alert the farmer in real time Keywords UWB tag RTLS Animal behaviour Behavioural indices Cow localisation Cow speed analysis Part I INTRODUCTION Chapter 1 PREFACE The increasing demand for livestock products such as meat eggs milk and derivatives represents an important issue in today s society and has signif icant implications on agricultural production methods Industrialisation of agricultural production in fact represents a direct consequence of this con cept by assuming an ever increasing scale directly linked to globalisation Intensive livestock farming i
92. ds 06 16 11 11 47 59 LocationEngineCell 0001 sensor proportion of slots with the wrong timeslot length 1000000ppm 06 16 11 11 47 59 LocationEngineCell 0001 sensor proportion of slots without a good set of phase samples 1000000 06 16 11 11 47 59 LocationEngineCell 0001 sensor proportion of slots without stable timing circuit phase 1000000ppm unknown bootstrap 00 11 CE 00 39 F5 192 168 0 100 2823 Jauto revc master elf block 400 unknown bootstrap 00 11 CE 00 39 F5 192 168 0 100 2823 Jauto revc master elf block 500 LocationEngineCell 0001 sensor timing cable error report for previous 15 seconds 06 16 11 11 48 06 LocationEngineCell 0001 sensor proportion of slots with the wrong timeslot length 1000000 06 16 11 11 48 06 LocationEngineCell 0001 sensor proportion of slots without a good set of phase samples 1000000ppm 06 16 11 11 48 06 LocationEngineCell 0001 sensor proportion of slots without stable timing circuit phase 1000000ppm_ 06 16 11 11 48 07 192 168 0 105 unknown 00 11 CE 00 39 F5 192 168 0 100 2823 Jauto revc master elf block 600 06 16 11 11 48 10 192 168 0 105 unknown 00 39 5 192 168 0 100 2823 auto revc master elf block 700 06 16 11 11 48 13 192 168 0 105 unknown 39 5 192 168 0 100 2823 Jauto_revc_master elf block 800 06 16 11 11 48 16 192 168 0 105 unknown 39 F5 192 168 0 100 2823 Jauto_revc_master elf block 900 06 16 11 11 48 19 192 168 0 105 unknown 39
93. e following relation ds d dx 2 dy 2 E 6 6 WA d od 60 The instantSpeed2D object was allocated in order to contain the instan taneous speeds 1 List lt ScalarVelocityNDEvent gt instantSpeed2D new List ScalarVelocity NDEvent gt Equation 6 6 was implemented in the following code ScalarVelocityNDEvent instantSpeedND2 new ScalarVelocityNDEvent meanData i MiddleDate meanData i IdTag Math Sqrt Math Pow meanData i X meanData i 6 5 SOFTWARE TOOL FOR THE VISUAL ANALYSIS 109 1 X window 2 Math Pow meanData i Y meanData 1 1 Y window 2 where window is a variable containing the number of seconds of the sampling instant set In our case window 5 After the instantaneous speeds are calculated an X MLParser was im plemented in order to get a file with extension Xml containing the data to be plotted which are related to the computed velocities Here is the code of the method that implements this function 1 2 private void WriteXmlWithVelocityData XmlTextWriter writer new XmlTextWriter C data xml null 5 writer Formatting Formatting Indented 6 writer WriteStartDocument 7 8 open plot tag with id 2 9 writer WriteStartElement plot 10 writer WriteAttributeString id 2 1 12 open curves tag 13 writer WriteSta
94. e to organize these folders a software tool written in C language was developed which using the recursive algorithm quicksort Demetrescu et al 2004 performed first a sorting by date of acquisition of the images names and then returned in output a text file for each day of acquisition where the identifier in each row of the current image is followed by the acquisition date in the following format ID IMAGE YYYY MM DD H M S Below is the code used in the implementation of the algorithm quicksort contained in the specifically developed tool void quicksort array fileStructType gt a int left int right if left right NAS 3 called FileIndex_ yyyy _ mm _ dd txt 6 4 RTLS PERFORMANCES 91 int pivot partition a left right quicksort a left pivot 1 quicksort a pivot 1 right int partition array fileStructType gt a int left int right int64 pivot a left date while true while a left date lt pivot left 4 while a right date gt pivot right if left lt right String tempName int64 tempDate tempName a left FileName tempDate a left date a left FileName a right FileName a left date a right date a right FileName tempName a right date tempDate else return right Importing data on database Once the folder containing all the index files g
95. e fixed with mounting brackets supplied by Ubisense on the metal beams that support the roof of the barn through plastic cable ties spirit level was used across the back of the sensors case to ensure there was no roll and the pitch and yaw angles were calculated in order to direct the sensors towards the floor in the middle of the area under study Once fixed their position thin films and flexible plastic material were used to entirely cover the sensors in order to protect them from the surrounding environment Figure 6 11 c Figure 6 11 Sensors IP30 Serie 7000 used in the trial a front side of a sensor b back side of Master sensor with Timing Cables black and power cable white c sensor covered with protective plastic material A wired network consisting of CAT5 cables and RJ45 modular connector was designed and installed to the support structure of the barn roof in order to connect the sensors to the PC via the PoE switch This network allowed both the power of the sensors and the data transmission between sensors and PC Another network was installed in order to connect the sensors to each other through the Timing Cables CAT5e cables to allow their temporal synchronization by calculating the TDOA between each pair of sensors As 66 CHAPTER 6 THE CASE STUDY already mentioned in section 5 3 2 the Ubisense system is composed of two types of sensors Master and Slave Although Master and Slave sensors a
96. e old row Location RowType new row ANALYSIS AND INSTALLATION OF UBISENSE RTLS 75 3 if this InvokeRequired 4 this BeginInvoke new MethodInvoker delegate CellData Update old row new row 5 else 6 LocationOutput new row 7 j Output of the location of the object within Location Row Type r 1 private void LocationOutput Location RowType r string id r object Id ToString Substring 15 4 string type r object_ DynamicType Name 5 string stdErr r accuracy stderr ToString 6 double x Math Round r position P X 3 7 double Math Round r position P Y 3 8 double z Math Round r position P Z 3 9 string datetime r time_ ToLocalTime ToString r time_ Millisecond 10 object values datetime id x z type 1 StringBuilder sb new StringBuilder datetime 12 13 sb Append Append id Append Append x Append Append y Append Append z Append Append stdErr Append 14 15 log WriteLine sb ToString 16 log Flush IY 18 insertInList id 19 Acquisition of the name of the given object 1 static protected string object name UObject obj 3 if naming schema null 4 using Ubisense UName Naming ReadTransaction xact naming schema ReadTransaction 5 foreach Ubisense UName Naming ObjectName RowType row in Ubisense UName Naming ObjectName object name xact obj
97. e tag XYZ and cell boxes show the tag id its co ordinates and the cell id The check boxes at the bottom determine what information about events is shown in the Cell Map The All and None buttons select all check boxes and no check boxes respectively 6 2 5 Data storage The software provided by Ubisense however does not allow the disk storage of the acquired data from the RTLS For this reason a special software was developed As already introduced in section 5 3 1 3 you can create software applications interfacing directly with the Ubisense Location Platform and make calls to all services through the use of Ubisense NET APIs According to this consideration using Microsoft R Visual Studio a free programming environment distributed by Microsoft R Net framework it was possible to automate the necessary operations for data storage on disk so that to they could subsequently be processed This tool which was running on the virtual machine where it was installed the software provided by Ubisense by executing a connection to the Ubisense Location Platform at 1 s interval requested the latest data acquired by the sensors and wrote the following information to a file with csv extension acquisition date acquisition time tag identification number ID location in space expressed by the coordinates x y and z standard error The main instructions used for the implementation of the software tool are listed below CHAPT
98. e the following hardware Ubisense hardware components as seen in 5 3 1 1 section server to send and receive location information A PoE enabled network switch and CAT5e cables to build the sensors network A DHCP server to assign the desired IP addresses to the sensors An internet connection The following software requirements should be available to run in a stable Microsoft R operative system Ubisense software components 5 3 1 2 Software applications 5 3 1 3 DHCP server 6 2 2 Analysis of the RTLS functionality A detailed study on the RTLS functionality implemented in the case study in question was carried out before the beginning of the system installation step In particular an analysis of Visual Modeling was carried out by adopting the Unified Modeling Language UML The Visual Modeling is a process that provides a graphic representation of a model using a well defined set of 54 CHAPTER 6 THE CASE STUDY graphical elements that depending on the used modeling language constitute the main components for the representation of particular diagrams In this study as already mentioned the UML was used i e a graphical language used in object oriented development that includes several types of system models able to provide different views of a system The UML is the standard for object oriented modeling Sommerville 2010 In the following sections some UML diagrams produced for the RTLS functionality a
99. e was the following Intel R Core 2 Quad CPU Q6700 2 66 Ghz processor and operating sys tem Windows R XP Professional 2002 SP3 Once installed the DHCP Server assigned to the sensors a pool of static IP addresses manually set each of them was uniquely associated to the MAC address of a sensor in order to separately send input commands to each sensor and recognize the sender of the messages received At this point referring to the user manual provided by Ubisense Ubisense 2008 2010 the suite of software components Ubisense Location Platform which is described in section 5 3 1 2 was installed Once completed the in stallation wizard the sensors which were initialized by the software using the Configuration Engine Location LEC application were booted During this operation a log window showed the messages issued by the system as a result of any transaction carried out in order to communicate the details of the communication with the sensors to the user Figure 6 13 However in Figure 6 14 is shown the status of the sensors after the boot operation The table in the figure shows the MAC address of the sensors the cell to which they belong the current status date and time in which they were active before the last update the IP address and any flags respectively Subsequently a power threshold that is related to the signal received by the sensors was set in the LEC If the received signal strength fell below the threshold it
100. e year study of lying and standing be haviour of dairy cows in a frestall barn in italy Journal of agricultural engineering 40 27 34 Rabinovich S G 2010 Evaluating measurement accuracy Springerverlag New York Reiners K Hegger A Hessel E B ck S Wendl G amp Weghe H Van den 2009 Application of rfid technology using passive hf transponders for the individual identification of weaned piglets at the feed trough Computers and Electronics in Agriculture 68 178 184 Roelofs J Eerdenburg F van Soede N amp Kemp B 2005 Pedometer readings for estrous detection and as predictor for time of ovulation in dairy cattle Theriogenology 64 1690 1703 Rorie R Bilby T amp Lester T 2002 Application of electronic estrus detection technologies to reproductive management of cattle Theri ogenology 57 137 148 Ruiz Garcia L amp Lunadei L 2011 The role of rfid in agriculture Ap plications limitations and challenges Comput Electron Agric 79 42 50 Sahinoglu Z Gezici S amp Guvenc I 2008 Ultra wideband positioning systems Vol 2 Cambridge university press Cambridge UK Samad A Murdeshwar P amp Hameed Z 2010 High credibility rfid based animal data recording system suitable for small holding rural dairy farmers Computers and Electronics in Agriculture 73 213 218 Scalera A Conzon D Brizzi P Tomasi R Spirito A amp Mertens 2013
101. enerated in the previous step was obtained a database was created by using the software Microsoft R Office Access which was populated by inserting all the records belonging to those files Therefore delete queries for duplicate data and sorting filters by date were carried out Exporting data from database Database was exported to a number of text files equal to the number of days of image recordings by using the tool already described in section 6 4 1 1 suitably adapted to the dataset of images This application performed an initial connection to the database and then a query for the selection of 92 CHAPTER 6 THE CASE STUDY records which were gradually written to a text file The structure of the records stored in these files is the following ID IMAGE YYYY MM DD M SS In Figure 6 33 is shown the structure of the dataset of images obtained after the execution of this software 15 08 2011 534839 2011 08 1560 1 534840 2011 08 15603 534841 2011 08 15605 534842 2011 08 1560 7 534843 2011 08 15609 534844 2011 08 156 0 11 un HUI mm Wm d RH M 534845 2011 08 1560 13 CC IRR TT 534846 2011 08 156015 534847 2011 08 15 6 0 17 IR 534848 2011 08 156019 534849 2011
102. ent be haviours for instance lying from perching with regard to the computation of cow lying index In addition an automatic and real time software tool for the visual anal ysis of the cows location data acquired by the RTLS principally based on the implementation of two tools the CowHeatMap and the CowSpeedGraph was developed This software uses the RTLS potentiality to provide the user with a useful tool in order to perform specific analyses such as the estrus detection Promising results with regard to both the performance of the im plemented features and their use were obtained in a use case when useful information related to the occurrence of the physiological state of estrus were automatically achieved This test showed that a pattern related to the behav ior of the analysed cow when the state of estrus occurred can be identified both in CowHeatMap and in CowSpeedGraph The use cases carried out in this study however was only one way to test the functionality of the software applied to real data to obtain significant achievements about the use of the tool in order to find useful information re lated to the occurrence of the estrus physiological state It is clear that a more detailed study carried out on a dataset composed of a greater number of data concerning a larger period and a greater number of cows is recommended in order to ensure the scalability and the effectiveness of the software The results obtained in this test
103. erable to linear or otherwise constricted layouts Where possible it is desirable to incorporate spatial variability in the z direction height of the receiver locations as well The position of sensor Sensso was chosen as origin of the Cartesian coordinate system used in the real environment EX ng t 1 o Sens y Sens of i d 15 10 xi Feeding passage 1 1 l T Tag rif i A 9 1 Feeding alley 1 1 1 d 10 00 i Stall row n 1 i g amp 1 amp 4 30 o 8 1 a A Stall row n 2 h forg 0 70 Service alley Sens Sens Section A A li SER Sensor IP30 Sensor IP30 b a Figure 6 10 Plan a and section b of the area of the barn under study with sensors layout 6 2 3 3 Hardware installation At this point of the installation process it was possible to perform the phys ical installation of the hardware components of the RTLS The following devices were used in this case study 4 Sensors IP30 Serie 7000 10 Compact Tag IP65 ANALYSIS AND INSTALLATION OF UBISENSE RTLS 65 1 PoE enabled network switch 1 PC with an Intel R Core 2 Quad CPU Q6700 2 66 Ghz processor and operating system Windows Vista R Business The sensors wer
104. erapies adopted and tend to increase in general the rate of un productive animals or even mortality rate of the herd New methods that aim at eliminating the use of such substances have 4 CHAPTER 1 PREFACE been recently introduced in order to reduce costs related to the treatment of animal diseases and adoption of antibiotics to accelerate their growth The development of new approaches based on the use of automated moni toring systems has featured a highly technological breakthrough capable of ensuring a growing condition of animal welfare and therefore food safety for consumers Tullo et al 2013 Within these new approaches Precision Livestock Farming PLF is a technology that includes and promotes techniques based on sensors con tactless sensing with image analysis or sound analysis wireless data trans mission traceability techniques etc to make smart farming a reality and create added value for many stakeholders animals farmers veterinarians feed and product suppliers health services policy makers the media and of course the consumer Berckmans 2013 According to Wathes 2010 PLF relies on four essential elements 1 the continuous sensing of the process responses at an appropriate fre quency and scale with a continuous exchange of information with the process controller 2 a compact mathematical model which predicts the dynamic responses of each process output to variation of the inputs and can be and is
105. escribed in section 5 2 1 a careful analysis is recommended with regard to the physical features of the location and the indeed use of the system in order to determine the possible constraints to be taken into account in the next phases Subsequently the design of the sensors layout through the preliminary definition of their geometrical arrangement is recommended This step is crucial to get the optimal performance of the system because if done inaccu rately it could compromise the proper functioning of the RTLS For instance if you have four sensors a simple configuration is a square with sides in the range of 10 30 m The sensors should be mounted at the corners of the square near the ceiling for good line of sight across the tracked space After defining the layout of the sensors a PC that performs the functions of the server for the RTLS must be configured To do this you need to install the software provided by Ubisense and a DHCP server as indicated in the reference guide Ubisense 2008 2010 After that it s possible to perform the physical installation of the sensors as determined in the previous step The sensors should point towards the floor in the middle of the space and should have no roll In addition it is necessary to connect the sensors to the server PC through a PoE enabled network switch Master and Slave sensors are physically the same The minimum requirement for the network cables is CAT5e Power over Ether
106. essment of the RTLS the installation of a multi camera video recording system was carried out to acquire the panoramic top view images of the barn area under study Furthermore specific software were implemented to allow building and linking of two datasets composed of the panoramic images obtained from video recordings and the tag positions mea sured by the RTLS as well as allow visual recognition of tag position within the panoramic images This software allows an interactive accurate and au tomatic selection and control of the datasets as well as image visualization and labelling with the aim of the computation of the true tag positions Localisation and identification performances of RTLS were then computed by using suitable indices and metrics also based on an outlier data cleaning technique On this basis tag positions and identification errors as well as the suitable trade off between performances were provided Significant improve ments of the results were achieved as the average localisation error reduced of about 0 046 m in dynamic conditions Trade off of RTLS performances 163 164 CHAPTER 9 CONCLUSIONS yielded an average localisation error equal to about 0 52 m with a position accuracy of 9896 for cows tag and an error of about 0 11 m with a nearly 10096 accuracy for the reference tag Further improvement of this research could regard the evaluation of the localisation error along the z direction in order to distinguish differ
107. eviation whereas RTLS identification performance was assessed by computing precision and sensitivity The precision of a mea surement system also called reproducibility or repeatability is the degree to which repeated measurements under unchanged conditions show the same results It was calculated through the relation number of TP number of TP number of FP precision 6 2 where TP True Positive and FP False Positive The sensitivity instead also called recall or true positive rate measures the proportion of actual positives which are correctly identified as such It was calculated through the following relation 6 4 RTLS PERFORMANCES 97 number of TP number of TP number of FN sensitivity 6 3 where FN False Negative For each tag the number of true positives was obtained by counting the number of times that the cow was successfully detected in its position by the RTLS the number of false positives was obtained by counting the number of times that the cow was wrongly detected in its position by the RTLS and the number of false negatives was obtained by counting the number of times that the cow was not detected by the RTLS though present in the framed scene Finally three performance metrics Metric A Metric B and Metric C were defined to establish the trade off between the localisation and identifi cation performances of the RTLS Metric A for each tag all the were used to compute the mean l
108. ext the real time identification and localisation of each dairy cows of the herd play an important role and several studies in the field of precision livestock farming have been carried out 4 1 Animal identi cation With regard to cow identification Rorie et al 2002 studied the application of commercially available electronic estrus detection technologies to repro ductive management of cattle and demonstrated that pedometers are the most applicable devices to lactating dairy cattle and have a higher accu racy and efficiency when combined with visual observation Also Roelofs et al 2005 adopted a pedometer for oestrous detection as predictor for time ovulation in dairy cattle and demonstrated that pedometers can accurately detect estrus though not in real time The collection of estrus data in real time is of considerable importance to avoid delayed inseminations of cows which could reduce cow fertilization rate and as a consequence have negative economic impacts on farm budget and costs To this aim recent research Brehme et al 2008 J nsson et al 2011 achieved excellent results by adopting real time pedometers for estrus 13 14 CHAPTER 4 ICT APPLICATIONS detection in dairy cows However pedometers do not allow cow localisation and cannot be used to distinguish some behavioural patterns e g standing vs feeding perching that are crucial data as it has been proved that the daily incidence of standing behaviour is of co
109. f Applied Geodesy 4 23 Oppermann I H m l inen M amp Iinatti J 2005 theory and applications John Wiley amp Sons Ltd The Atrium Southern Gate Chichester West Sussex PO 19 850 England Oshana R 2006 Dsp software development techniques for embedded and real time systems Elsevier Ed 30 Corporate Drive Suite 400 Burlington MA 01803 USA Linacre House Jordan Hill Oxford OX2 8DP UK Newnes Overton M Sischo W Temple G amp Moore D 2002 Using time lapse video photography to assess dairy cattle lying behavior in a free stall barn Journal of Dairy Science 85 2407 2413 Pastell M Tiusanen J Hakojarvi M amp Hanninen L 2009 A wireless accelerometer system with wavelet analysis for assessing lameness in cattle Biosystems Engineering 104 545 551 Peck R Olsen C amp Devore J 2011 Introduction to statistics and data analysis CengageBrain com Boston MA 02210 USA Porto S Arcidiacono C Anguzza U amp Cascone G 2013 A computer vision based system for the automatic detection of lying behaviour of dairy cows in free stall barns Biosystems Engineering 115 184 194 168 References Porto S Arcidiacono C Cascone G Anguzza U Barbari M amp Si monini S 2012 Validation of an active rfid based system to detect pigs housed in pens Journal of Food Agriculture amp Environment 10 468 472 Provolo G amp Riva E 2009 On
110. f pan and zoom operation University of Catania Department of Agri food and Environmental Systems Management Les 160 Day Select m TAG 23 TAG 239 2001 08 01 Hide Open x y Y x y TAG 53 TAG 26 TAG 04 Contr SSS me m Goto Curent Postion 8586 26827 cesa lt 1 e 50 3 TAG 20 TAG 51 Q TOE ames LI TrackBar Position 8586 Date 1 8 2011 10483 TU TE EE D 251438 Tags Data identikit mee Y Y 2 10439 1818 M 8933 4107 0 38 0 000157312359078787 2 0 000261390639934689 Date 01 08 2011 10 48 03 01 08 2011 10 48 03 TAG 254 9305 1053 093 VB 554 10285 093 Bi 0 00717613875772804 B 0 0020701892208308 Date 01 08 2011 10 48 03 01 08 2011 104801 23 TAG 08 x Y 2 Y z er VI 462 89M 0268 29 00000135968060931191 Date Date 01 08 2011 10 48 03 mam Y z Y z WM ass 129755 128 8753 11581 0892 Sd Std 8 0000581809203140438 84 0000897705496754497 Date 01 08 2011 10 48 06 01 08 2011 10 48 03 TAGS3 y TAGS y z 78 s 309 osa Bd 000586813455447555 Date 01 08 2011 10 48 03 Figure 7 11 An example of the computation of the true position of the tags with ID 004 and 008 7 1 RTLS PERFORMANCES 135 7 1 4 Planimetric position errors obtained by the RT LS In this section the planimetric position errors obtained by the R
111. f the graph For instance by always setting ymin 0 and y the area under the curve plotted will be highlighted in the graph In addition there is a label and a description At this point of the process data were ready to be plotted and then were displayed for the client in the form of speed graph To this end two data structures are implemented through the Point and Curves classes In particular the Point class which has the following source code was implemented to contain the coordinates and y to be plotted where x is the value of the instantaneous velocity and y is relative to the corresponding sampling instant N O o N 09 OC RB Q NY 0o Oo OC public class Point region Constructors and Destructors public Point DateTime x float y public Point string x float y X xi Y y public Point DateTime x float y float ymin float ymax public Point string float y float ymin float ymaz this x y YMin ymin YMax ymax endregion 112 CHAPTER 6 THE CASE STUDY 21 22 region Public Properties 23 24 public DateTime X get private set 25 public string X get private set 26 public float Y get private set 27 public float YMin get private set 28 public float YMax get private set 29 30 endregion 31 The class Curve which has the source code shown below was imple
112. file with dll extension present in one of the folder to install the software Ubisense and click OK Links to the references in the project in order to be able to use the classes implemented in these libraries will be automatically created Once you have set the project you can implement the desired functions The Ubisense NET API Reference Manual available in Visual Studio help if you have installed the Ubisense software is the golden reference for any detailed information about the API Ubisense company developed a number of software applications such as Visible Industrial Process Asset Manager Transit Yard Manager etc by offering innovative solutions to prestigious customers in different sectors in dustry manufacturing and automotive logistics military etc Heathcote 2011 Finally as already seen in Figure 5 9 it is possible to interface various types of Corporate Systems e g ERP MES PPS WMS and DMS with the Ubisense Location Platform and or with specifically created software applications in order to integrate the system Ubisense in an information technology management system called information system in information technology and place it in relation to other relevant processes which are already available in it 5 3 2 Operating speci cations In the previous sections the features of the main components hardware and software of the Ubisense system were described 5 3 UBISENSE UWB RTLS 43 In this
113. for the fotu He zo s Ld qom Gone ee 140 RTLS identification and localisation performances computed for each one of the eight tags analysed and for the reference tag ID 187 for the uunc somom mg oos Behe de xw 141 xiii xiv LIST OF TABLES LIST OF TABLES Xv Abstract The main objective of this study was to evaluate the localisation and iden tification performances of a Real Time Location System RTLS based on Ultra Wide Band UWB technology within a free stall barn which repre sents a particularly harsh environment for the functioning of this kind of system Each dairy cow was equipped with an active tag applied to one ear video recording system was installed in the barn to perform the assessment of the RTLS Top view camera images of the area of the barn under study were rectified and synchronised with the RTLS Each position of the cow computed by the RTLS was validated by performing cow visual recognition on the camera images To perform this validation a software specifically designed for the purpose was utilized It is a quick accurate automatic and interactive tool which includes selection and control tabs for data management visualization and labelling of the images with the aim of the computation of the tag true positions Localisation and identification performances of the RTLS were assessed by applying an outlier data cleaning technique to tag localisation er
114. for which one of the first two criteria described above was true 6 4 RTLS PERFORMANCES 95 The structure of the records obtained after the matching operation is the following ID IMAGE YYYY IMAGE M IMAGE D IMAGE H IMAGE M IMAGE S IMAGE YYYY TAG MM TAG DD TAG H TAG MM TAG SS TAG ID TAG X Y Z STD ERROR The text files returned in output to the end of this phase presented the format shown in Figure 6 35 430456 2011 430457 2011 430459 2011 430459 2011 630460 2011 6304601 2011 432462 22211 430463 2011 430464 2011 430465 2011 430466 2211 430467 2034 630463 2211 490468 2011 630470 2011 490472 2011 430472 2011 430473 2011 430474 2911 30475 20211 2011 39511 820009086004 7 24 0 56 2 839 2 00968237593770027 020000006004 6 347 1 549 0 923 0 0104614170260170 020000096004 6 349 0 92 1 236 0 0437256144664283 020000086004 6 681 0 711 0 182 0 00912768952542504 020000086004 6 845 0 561 1 551 0 0061085224256672 020000086004 7 927 3 237 1 314 95 0041033473403832 020000086004 7 08 1 645 0 117 0 0392231222316254 020000086004 6 919 1 094 1 978 0 00155301799532026 020000086004 7 669 1 652 0 47 5 05185144669376314 020000086004 7 649 1 653 0 47 0 00185244669376311 020000086004 7 924 1 979 0 4 7 0 01948763 7684755 0200000840904 7 924 1 979 5 457 0 0194427 6957 04755 030000086004 7 701 1 385 1 385 0 00738782902347457 020000004004 7 377 1 923 1 379 0 000 2771 1701 6152 020000086004 7 504 1 67 0 48 O 003997602965682
115. form simul taneously locating operations and data transmission in communication systems usually involves the presence of numerous collisions between the two traffic flows with the result to obtain a low bit rate and a high imprecision in the estimation of distances The use of an extremely wide bandwidth as that of UWB make it possible to divide the traffic flows properly and avoid collisions Di Noia 2010 Security the very low level of power emitted by the tags delivers high security within a system that uses UWB technology A UWB tag transmits in fact at a power level of 10000 times lower than a typical mobile phone Domain 2013 No interference with other devices the low power and wide range of frequencies used for a UWB signal ensures the almost complete 5 1 ULTRA WIDEBAND SYSTEMS 21 absence of interference with other wireless technologies such as Wi Fi Bluetooth UHF or handheld passive readers PDA passive read ers Figure 5 1 Tests showed that the interference impact on UWB from a wireless technology like Wi Fi is essentially negligible lt 0 3 Domain 2013 RFID SPECTRUM WHA BLUETOOTH POWER LEVEL milliWatts 2 000 2 0 4995 p CELL PHONE BLUETOOTH HEADSET UWBTAG Figure 5 1 Frequency spectrum and power level of UWB signal in comparison to other wireless technologies Domain 2013 A system based on UWB technology is therefore characterized by h
116. future there will be a large class of objects natively interconnected which will allow for new uses including new ways of working new ways of entertainment and animation new ways of living and last but not least important new ways of farming A recently emerging computing concept namely Internet of Things IoT could be included as part of this scenario bypassing the interoperability problems often typical of highly specialized proprietary devices which are generally designed to be integrated into systems that are extremely closed themselves This new approach considers all sources of information RFID tags Real Time Location Systems sensors actuators mobile phones etc such as assets which are uniquely indexed by an address assigned by a centralized system that can interact and cooperate with each of them to achieve common goals Scalera et al 2013 CHAPTER 1 PREFACE Chapter 2 OBJECTIVE OF THE STUDY In the research described in this thesis work a Real Time Location System RTLS based on the UWB technology was utilised for the identification and localisation of a group of dairy cows housed in a free stall barn The RTLS used in this work Ubisense UK is currently recognised to provide one of the highest accuracies of all the RTLSs Linde 2006 Weichert et al 2010 The achievable accuracy which Ubisense declares for the localisation of moving objects in real time is 15 cm in the three dimensions x y z i e
117. fy the effectiveness of the results obtained from the use of the tool in order to 118 CHAPTER 6 THE CASE STUDY find useful informations related to the occurrence of the physiological state of estrus The date analysed in this test were acquired by the RTLS from 10th Au gust 2013 to 16th August 2013 In particular only the location data related to the tag with ID 020 were used as the cow which that tag was associated manifested the state of estrus visually observed by the farmer at around 9 30 am on 12th August 2013 For each of the selected days two time intervals were identified 1 2 CF Time Cow Feeding Time where the CFI Cow Feeding Index calculated as the ratio between the number of cows that are to the manger and the total number of cows present in the barn Equation 6 7 took on average the maximum values as shown in Figure 6 38 green area in the same period considered in this test and in the same livestock environment in a study carried out by Agosta 2012 cows in feeding CFI 6 7 tot cows In particular CFTime is composed of two time intervals the first be tween 06 00 and 09 30 and the second between 18 00 and 19 30 CLTime Cow Lying Time where the CLI Cow Lying Index defined as the ratio between the number of cows in decubitus posi tion inside of the bunks and the total number of cows present in the barn Equation 6 8 assumed on average maximum values Figure 6 38 yellow a
118. gewerf P 2013 Validation and application of an indoor localization system for animals In Precision livestock farming 2013 p 135 144 Leuven Belgium References 167 Jonsson R Blanke M Poulsen N Caponetti F amp Hojsgaard S 2011 Oestrus detection in dairy cows from activity and lying data using on line individual models Computers and Electronics in Agriculture 76 6 15 Kalman R E 1960 A new approach to linear filtering and prediction problems Transactions of the ASME Journal of Basic Engineering 82 35 45 Linde 2006 On aspects of indoor localization Unpublished mas ter s thesis Fakult t fur Elektro und Informationstechnik Universitat Dortmund Martinez Ortiz C Everson R amp Mottram T 2013 Video tracking of dairy cows for assessing mobility scores In D Berckmans amp J Vander meulen Eds Precision livestock farming 13 Leuven Belgium 10 12 september 13 Mattachini G Riva E amp Provolo G 2011 The lying and standing activity indices of dairy cows in free stall housing Applied Animal Behaviour Science 129 18 27 Mei 2003 Modeling and performance evaluation of a bppm uwb system Unpublished master s thesis Delft University of Technology Mok E Xia L Retscher G amp Tian H 2010 A case study on the feasibility and performance of an uwb aoa real time location system for resources management of civil construction projects Journal o
119. h feature fulfill from the point of view of the manufacturers the ability to realize inexpensive receivers Di Noia 2010 High bit rate bandwidth occupied by UWB systems allows the trans mission of large quantities of data in a given time interval bitrate reaching in short distances a speed greater than or equal to 0 5 Gbit s which is a decidedly high speed considering that it is a wireless telecom munications system Di Noia 2010 Immunity to multipath signals UWB pulses are pulses in broad band radio frequency of extremely limited duration from a few tens of picoseconds to a few nanoseconds This feature allows a better resolu tion of the multiple paths In fact due to the short pulse duration it is unlikely that a reflected signal interferes with a direct one Therefore it favours the identification and subsequent exclusion of the reflected signals This property of UWB signals is the main reason for which this technology is used by the Real Time Location Systems RTLS for locating the objects in closed environments or with high density of obstacles Di Noia 2010 Ability to penetrate through obstacles the absorption spectrum of various materials metals in particular is concentrated on narrow frequency bands The UWB signals as already seen are broadband signals that are not strongly attenuated by the obstacles unlike other narrow band signals Di Noia 2010 Localisation and communication simultaneously per
120. he RTLS detections On the basis of these considerations the best trade off between localisa tion and identification performances of the RTLS was obtained when adopt 7 2 SOFTWARE TOOL FOR THE VISUAL ANALYSIS 141 ing Metric B Regarding the reference tag precision and sensitivity resulted 1 for Metric A and 0 99 for Metric B and Metric C and the values of mean error standard deviation resulted 0 11 0 05 for all the metrics Tables 7 4 7 5 and 7 6 Therefore excellent results were obtained for both localisation and iden tification performances regardless of the metric adopted This mean error is in good agreement with the value of 0 1225 m declared by Ubisense Metric C Tag Identification Localisation ID Mean Error Precision Sensitivity Std m 004 0 42 0 42 0 18 0 20 008 0 61 0 61 0 12 0 19 020 0 61 0 61 0 17 0 20 023 0 11 0 11 0 31 0 20 026 0 32 0 32 0 24 0 20 053 0 53 0 53 0 20 0 21 239 0 61 0 61 0 15 0 20 254 0 54 0 54 0 20 0 20 187 0 99 0 99 0 11 0 05 Table 7 6 RTLS identification and localisation performances computed for each one of the eight tags analysed and for the reference tag ID 187 for the Metric C 7 2 THE AUTOMATIC AND REAL TIME SOFT WARE TOOL FOR THE VISUAL ANAL YSIS OF THE COWS LOCATION DATA 7 2 1 Features of the software As described in section 6 5 the software specially developed for the real time visual analysis of the cows location data in free stall barns was principally b
121. he two buttons in question would cause an error 130 CHAPTER 7 RESULTS Under the Control tab there is section dedicated to the display of informations of the current record In particular it is constituted by Image tab that shows Date and ID of the current image used for the visual recognition of the position of each tag Tags Data tab which shows for each tag X Y and Z coordinates Standard error and Date related to the current record Identikit tab which allows the visualization of two images of each cow associated to the identification number of the tag see Figure 7 7 Current Tot Tag textbox which shows the total number of tags asso ciated to the current image in the operation of matching between the two datasets Image Date 23 8 2011 10 31 1 CP Tp 1D 755010 Tags Data TAG 26 4 COW 155 Figure 7 7 Identikit tab of the Software tool It allows the visualization of two images of each cow associated to the identification number of the tag The section located on the right side of the UI is dedicated to the man agement of the commands suitable for the visual recognition of the position of each tag and any manual correction of that position 7 1 RTLS PERFORMANCES 131 In particular a graphical representation of the study area called Map which consists in the overlap of the top view panoramic image corresponding to the curren
122. he user in the interface section dedicated to cowSpeedGraph and to interpret the answers received 6 5 SOFTWARE TOOL FOR THE VISUAL ANALYSIS 117 As already introduced in section 6 5 3 1 the client receives from the server a file with extension Html containing the dataset to be processed for the construction of cowSpeedGraph This file is in the following format oo NO OC RR Q N w M lt html gt lt head gt lt script type text javascript src dygraph combined js gt lt script gt lt head gt lt body gt lt div id graphdiv gt lt div gt lt script type text javascript gt g new Dygraph containing div document getElementById graphdiv new Date 2013 04 17 6 16 03 0 0123 new Date 2013 04 17 6 16 08 0 0083 new Date 2013 04 17 6 16 13 0 0297 new Date 2013 04 17 6 16 18 0 0521 labels Date COW SPEED fillGraph true lt script gt lt body gt lt html gt The instructions just shown implement the functions that can be inter preted by a web browser for the construction of cowSpeedGraph In particular the fast flexible and open source JavaScript charting library called dygraph combined js Dygraph 2013 it is used 6 5 4 use case of the software tool A use case of the software tools described in this section was carried out in order to assess the performances of the implemented features and to veri
123. her parameters could be found in the Ubisense guide Ubisense 2008 2010 6 2 9 Cow localisation After the filter setting it was possible to carry out the operation of cow localisation During this phase by starting the LEC tool it was possible to monitor in real time all the data captured by the RTLS related to the cows tags and the reference tag 84 CHAPTER 6 THE CASE STUDY Figures 6 27 6 28 6 29 and 6 30 show screenshots of the LEC during this phase In Particular the Figure 6 27 shows the 2D view of the area of the test where the reference tag is represented by a graphic element with a shape of a red circle Green lines represent the angle of arrival AOA and blue curves represent the time difference of arrival TDOA The elements represented by the symbol are the points used for static test section 6 2 6 and for calibration which were manually entered into the software In Figure 6 28 instead in the 3D reproduction of the study area is shown and the displayed tag is one of the tags applied to the cows Figures 6 29 and 6 30 show instead the position of the reference tag estimated by the RTLS relative to reference tag which corresponds exactly to the point indicated by the symbol This point in turn corresponds to the point where was physically installed that tag measured in the real environment Figure 6 27 2D view of the object of the test in the LEC tool ANALYSIS AND INSTALLATION OF UBISE
124. iagram sequence diagram is a UML model that shows the sequence of interac tions required to compose some operations In particular the interactions between actors and the system and between system components are repre sented Sommerville 2010 The key idea in this type of diagram is that the interactions between the objects are carried out by following a precise order and that this sequence occurs in time from the start to the end The Se quence diagram consists of objects represented by rectangles with a name messages represented by continuous lines with an arrow at the end and the time represented as a vertical progression Objects are arranged in sequence from left to right From each rectangle a dotted line called lifeline is drawn downwards Along the lifeline there is a small rectangle called activation The activation is the execution of an operation to which the object is responsible The length of the rectangle however represents the duration of the activation A message that travels from one object to another is drawn from the lifeline of the object from which the message leaves to the lifeline of the object to which the message is directed It can also occur that an object sends a message to itself that is a message that starts from its lifeline and arrives to the same lifeline When viewed in the vertical direction the sequence diagram represents the flow of time Graphically the time starts at the base of each ob
125. igh bandwidth pulses greater than 500 MHz for communication and or for the localisation obtaining high transmission speed a high resolution of the mul tiple paths high robustness to interference with other wireless technologies and the presence of obstacles high security and low cost of implementation 5 1 1 Ultra Wideband signals According to the Federal Communications Commission FCC FCC 2002 an UWB signal is defined as any signal that has a fractional bandwidth greater than 0 2 or occupies 500 MHz or more of spectrum The main feature 22 CHAPTER 5 UWB RTLS of UWB signals is that they occupy a much wider frequency band than conventional signals hence they need to share the existing spectrum with incumbent systems 5 1 1 1 Signal bandwidth The absolute bandwidth is calculated as the difference between the upper frequency fp of the 10 dB emission point and the lower frequency fr of the 10 dB emission point B fg fr 5 1 which is also called 10 dB bandwidth Figure 5 2 On the other hand the fractional bandwidth is defined as B B 5 2 f f 5 2 where f is the center frequency and is given by eee 5 3 2 From 5 1 and 5 3 the fractional bandwidth in 5 2 can be expressed as Birac 2 fr 5 4 fa ft In conclusion as already mentioned before according to the FCC an UWB system with fe larger than 2 5 GHz must have an absolute bandwidth larger than 500 MHz and a UWB
126. in CLTime interval distinct by day 2 3 151 7 18 CowHeatMaps and CowSpeedGraphs relative to the tag with ID 020 in CLTime interval distinct by day 3 3 152 7 19 CowSpeedGraphs of to the tag with ID 020 distinct by day relative to hays 4 oos 6 ee eee RTE Ew 153 xii LIST OF FIGURES List of Tables 5 1 5 2 5 3 7 3 7 4 7 5 7 6 Ubisense sensors technical specification 0 0 39 Ubisense tags technical specification 36 Ubisense location platform specifications 40 Planimetric position errors computed on the data provided by the RTLS 136 Planimetric position errors computed on the data provided by the RTLS for each one of the eight tags analysed and for the reference tag ID 187 before outlier data cleaning 137 Planimetric position errors computed on the data provided by the RTLS for each one of the eight tags analysed and for the reference tag ID 187 after outlier data cleaning lt s soco cce n poene om predora t 137 RTLS identification and localisation performances computed for each one of the eight tags analysed and for the reference tag ID 187 for the Metric As coo s opor dox por ode RUE RR ORO barta X os 139 RTLS identification and localisation performances computed for each one of the eight tags analysed and for the reference tag ID 187
127. in section 6 4 1 3 The only differ ences between the output text files and the input ones are the values of the two coordinates r and y in the case where the operator makes and saves changes 134 University of Catania Department of Agri food and Environmental Systems Management CHAPTER 7 RESULTS EIE 57 23 239 2011 08 01 Hide Man x y x y x y TAG 26 TAG 04 Control A E y x y Goto Curent Postion 8586 26827 0 1 a o8 TAG 20 TAG 51 0 x y x y x y TrackBar Postion 8586 image Date 1 8 2011 10 48 3 s EE D 251438 Tags Data identit _ mas z We 10439 1818 V 8933 4107 075 88 0000157312359078787 89 01000261390639934689 Date 01 08 2011 104803 01 08 2011 10 48 03 TAG 254 2 3 2 YM 930 10863 0993 MM 554 10296 093 Bi 000117613875772804 B 010020701892208308 01 08 2011 10 48 03 01 08 2011 10 48 01 23 g x TAG 08 _ x Zr 462 8994 026 Sd 9 ber 82 0000135968060931191 Date Date 01 08 2011 10 48 03 4 x Y z VW asse 1295 1 MM 8753 11581 0892 2d 0000581809203740438 89 0 000897705496754497 Date 01 08 2011 104806 Dae 01 08 2011 10 48 03 y ELO VN so 302 054 B 000586813455447555 29 Date 01 08 2011104803 Figure 7 10 In Map section an example o
128. in the follow ing of the thesis The object to be localised transmitter is defined as the target node and the receivers as reference nodes Angle of Arrival AoA An measurement provides information about the direction of a signal sent from a target node and received by a reference node particularly about the angle between the two nodes as shown in Figure 5 6 Figure 5 6 Definition of AoA between two nodes the reference node black node mea sures the AoA by determining the angle v between itself and the target node gray node Commonly antenna arrays are employed in order to measure the AoA of a signal The information about the angle is obtained at an antenna array by measuring the differences in arrival times of an incoming signal at different antenna elements An example is illustrated in Figure 5 7 for AoA estimation at a uniform linear array ULA When the distance between the target and the reference nodes is suffi ciently large the incoming signal can be modeled as a planar wave front This results in sin 1 c seconds difference between the arrival times at consecu tive array elements where l is the inter element spacing 1 is the AoA and c represents the speed of light T herefore estimation of the time differences of arrivals provides angle information More advanced array structures such as uniform circular arrays UCAs and rectangular lattices operate on the same basic principle as the ULA For a nar
129. ion myAccessConn new OleDbConnection strAccessConn catch Exception ex 22 31 32 33 34 35 36 37 6 4 RTLS PERFORMANCES 89 Console WriteLine Error Failed to create a database connection n 0 ex Message return try OleDbCommand myAccessCommand new OleDbCommand strAccessSelect myAccessConn OleDbDataAdapter myDataAdapter new OleDbDataAdapter myAccessCommand Opening database connection myAccessConn Open myDataAdapter Fill myDataSet TagTable catch Exception ex Console WriteLine Error Failed to retrieve the required data from the DataBase n 0 ex Message return finally closing database connection myAccessConn Close Therefore the selected records were written to a text file by using a StreamWriter The structure of the records within these files is the following DD MM YYYY HH MM SS ID TAG X Y 7 SID ERROR Reorganisation of the dataset In addition through the software tool it was possible to automate the partitioning of the data into groups each representing a particular tag Fi nally additional operation of data organization were automated creating subgroups each of them related to a day of recording In other words the dataset of the tag was constituted by a number of folders equal to the number of tags each of them containing a number of text files equal to the number of days of
130. ion Engine Configuration tool then the system takes readings of the AoA from the tag and yield the orientation of the sensor from the average of the measurements set This calibration step must be performed on each sensor in order to set their parameters of orientation pitch yaw and roll within the network in each of them The second phase of calibration also called Cable calibration is similar to orientation calibration except that a cable offset is between two sensors Therefore the calibration uses two sensors the sensor getting its cable offset and the reference sensor which is ideally the master Cable offsets are just some arbitrary numbers indicating the offset from the master so that the master achieve a null value For this reason this latter calibration phase unlike the previous one must be carried out only on each of the slave sensors At the end of the operations just described assuming that the sensors are disposed in such a way as to form a square and the tag used for the calibration has been placed at the point where they intersect its diagonals in the map of the Location Engine tool it will be noted that all sensors are oriented toward the reference point as shown in Figure 5 17 Figure 5 17 Map of sensor network after calibration operations Each sensor is oriented toward the calibration point the center where the tag used for calibration red dot is placed The green lines outgoing from each sensor focus the
131. ion aims at transforming the data in the correct format in order to perform a mapping of these on a panoramic image of the barn According to the pixel dimensions of the image the relations 6 8 and 6 7 are defined to get xtagPx and ytagPx i e the x coordinate value and the y coordinated value in pixels within the image respectively xtagPx 47 ztag k 6 4 ytagPx 613 ytag k 6 5 where xtag and ytag are x and y coordinates in meters respectively 35 67 This operation is iterated for each record automatically extracted Once the sequence of the couple xtagPzr ytagPy is obtained the soft ware executes a specially created script to get an output file with extension js containing the structure of the dataset to be provided as input to the client in order to create the HeatMap Below is the source code of the method that implements these steps 1 2 3 4 private void WriteJsWithLocationData string idCow LocationEvent locEvent string fileName C datasetHeatMap_ idCow Ja StringBuilder content new StringBuilder content AppendLine function global content AppendLine var events e locEvent Length content AppendLine var locations new Array events content AppendLine content AppendLine Array 2 for i 0 i events 1 locations content AppendLine ys content AppendLi
132. ject and continue to the bottom In Figure 6 7 the sequence diagram of the case study in question is shown The main objects are three Location Platform Sensor and Tags The Lo cation Platform is always active during the entire process of localisation in fact the activation of the Location Platform object has a height equal to that of the lifeline The first operation is carried out in chronological order from the Loca tion Platform object which sends the startSensors message to the object Sensor in order to activate it Once activated the object Sensor sends the interrogateTag RFSignal message to the Tag object to denote the oper ation of the activation request made by the sensor through an RF signal to the tag Once processed the received message the object Tag transmits 60 CHAPTER 6 THE CASE STUDY startSensors interrogateTag RFSignal respondSensors UWBPulse trackedTag t1 ti ol oii iz1 N N number of Sensors computeTDoA t1 ti updateAndStoreData TagLocation Figure 6 7 Sequence diagram ANALYSIS AND INSTALLATION OF UBISENSE RTLS 61 the respondSensors UWBPulse response message to the object Sensor to indicate the operation of UWB impulse transmission from the tags to the sensors At this point the activation relative to the object Tag is stopped since the tag task in the localisation process ends Therefore the next oper ation is carried
133. k out Shift drag will pan The main features of CowSpeedGraph are Handles huge data sets dygraphs plots millions of points without getting bogged down Interactive out of the box zoom pan and mouse over are on by default 7 2 SOFTWARE TOOL FOR THE VISUAL ANALYSIS 143 Dygraphs is highly compatible it works in all major browsers in cluding IES It is possible even to pinch in order to zoom on mo bile tablet devices 7 2 2 Software User Interface The UI of the software at boot time is shown in Figure 7 13 Real time software tool for the analysis of the cow s location data a 20000139187 lt ap UNIVERSITY yi W OF CATANIA DiGeS Figure 7 13 A screenshot of the software UI at boot time The main content of the UI is divided into two sections the controls section and the visualisation section The controls section contains two clickable elements that allow the selec tion of the type of function being performed CowHeatMap or CowSpeed Graph two text eld for entering tag ID and the number of back hours from now which can be chosen for carrying out the analysis and a setting panel through which you can set a variety of parameters related to the CowSpeed Graph display such as Show labels it lets you choose whether or not display the value on the top right corner of the point on the graph where the pointer is located 144 CHAPTER 7 RESULTS Grid line color
134. lic DateTime Date 5 public string IdTag 6 public double X 7 public double Y 8 public double Z 9 endregion 10 11 region constructors 12 internal ScalarVelocityEvent 13 DateTime date 14 string idTag 15 double x 16 double y 17 double z 18 1 19 this Date date 20 this IdTag idTag 21 this X x 22 this Y y 23 this Z 2 24 25 7 fendregion The ScalarVelocityNDEvent class finally was implemented to contain 108 CHAPTER 6 THE CASE STUDY the date the ID of the tag and the instantaneous velocity with respect to that instant of sampling The following source code refers to the class just described y O OC A WN O o WANDA A UN 20 public class ScalarVelocityNDEvent region properties public DateTime Date public string IdTag public double InstantSpeed 7 endregion region constructors internal ScalarVelocityNDEvent DateTime date string idTag double instantSpeed this Date date this IdTag idTag this InstantSpeed instantSpeed endregion At this point of the process a sampling of the selected data was per formed In particular a sampling time of 5 seconds was chosen in order to calculate the average positions for each interval in the three coordinates y and z Once the sampling was performed instantaneous speeds in two dimen sions y for each sampling instant i were calculated according to th
135. lic curves represent the hyperbolic localisation TDOA Red dot is the tag position Heathcote 2011 5 3 UBISENSE UWB RTLS 45 5 3 3 Installation procedure The installation phase is an activity which requires a lot of care and atten tion installation carried out inaccurately without following the recom mendations provided by Ubisense could lead to malfunctioning or alter the performance of the system This section describes the main steps of the installation process gener alized to all case studies More detailed information could be found in the online Ubisense How To articles Ubisense 2008 2010 To develop the system a path divided into five steps should be followed as shown in Figure 5 16 Figure 5 16 Installation Procedure of Ubisense system structured in five steps Ward 2010 Referring to Figure 5 9 the installation procedure described below will regard the hardware and middleware of the Ubisense platform since they constitute the basic elements for the development of the system The software interfacing with other systems or applications which is specifically developed for specific objectives will not be discussed in this section because it depend on the case study which the system is applied to 46 CHAPTER 5 UWB RTLS In the initial step of the installation process a survey should be performed to verify the environmental features of the space where you want to imple ment the localisation system As d
136. lisation In a typical open environment a location accuracy of about 15cm can be achieved across 95 of readings Scalable real time performance The Ubisense RTLS is designed to support multiple integrated real time location applications that work over large areas for large numbers of users Development and deployment tools The Ubisense platform pro vides a number of visual configuration and development tools which simplify the generation of end user applications Steggles amp Gschwind 2005 5 3 UBISENSE UWB RTLS 33 Furthermore Ubisense platform allows any location or sensor system to be plugged into its platform protecting users being locked to a single vendor This reduces the cost of ownership Ubisense platform architecture consists of three interconnected main ers i hardware components ii software middleware and iii software ap plications as shown in Figure 5 9 Heathcote 2011 00152152 P zi efe fff Sula CJ CESTA 70 22 20 m u n n Ut lt Un jen ope Aen mv Teh Ubisense Sensor Network Optional additional receivers and sensors Fw M e Oro VM a c aj RFID radiation temperature Hardware Figure 5 9 Ubisense platform architecture Hardware level contains the Ubisense Sen sor Network and optional receivers and sensors Middleware level contains the Ubisense Location Platform
137. mmand ExecuteNonQuery conn Close The CreateTableFromSelected TagBy TimePeriod method takes as input a string containing the ID of the tag and a numerical value corresponding to the number of hours to be subtracted from the current time and performs a SQL SELECT query Through this query the rows in the id Tag table which have in the Dates field a date that belongs to the interval h x h where x is the parameter corresponding to the hours passed in the input and h is the current time are selected 6 5 2 CowHeatMap implementation As it is shown in the flow chart of Figure 6 37 the implementation of the CowHeatMap was managed through two separate modules the server side and the client side Sections 6 5 2 1 and 6 5 2 2 describe the operations performed by the in dividual modules 6 5 2 1 Server side The server side software module was designed with the purpose of providing services to the client side module implemented in order to satisfy the requests received In particular after receiving a request from the client side module which ask to view the HeatMap relative to the data selected in the previous section 6 5 1 a conversion from meters to pixels of the coordinates z and y is performed on the data extracted from the database with the following 6 5 SOFTWARE TOOL FOR THE VISUAL ANALYSIS 103 format acquisition date acquisition time tag ID x y z standard error This operat
138. nalysis will be analyzed individually use case diagram section 6 2 2 1 class diagram section 6 2 2 2 state diagram section 6 2 2 3 and sequence diagram section 6 2 2 4 Since it comes to high level diagrams therefore all the features of the properties and operations related to the system will not be described but the most important steps that constitute the process of localisation of the RTLS will analysed without going into details 6 2 2 1 Use case diagram use case diagram shows the interactions between a system and its environ ment Use cases are collections of scenarios involving the use of the system where each scenario describes a sequence of events The main element of a use case diagram is the actor i e the entity that interacts with a use case by starting the sequence of actions described by the use case itself and pos sibly getting precise answers from the system It can be a person or another system Graphically as shown in Figure 6 4 an ellipse represents a use case and a little man is an actor The name of the actor appears just below the representation of the same actor as the name of the use case appears inside the ellipse An association line finally connects an actor to the use case and represents the communication between the actor and the use case In this case study the actor is represented from the User who has the task of assigning tags to the cows to locate It also has the task to activate the tags
139. ne locations double x double y for int i 0 i lt locEvent Length 1 i x 47 locEvent i X 35 67 18 19 20 21 22 23 24 25 26 27 28 29 30 104 CHAPTER 6 THE CASE STUDY y 613 locEvent i Y 35 67 content AppendLine pU ae x ToString Replace y ToString Replace x p 77 locEvent i Date ToString x 47 locEvent locEvent Length 1 DX 35 67 y 613 locEvent locEvent Length 1 Y 35 67 content AppendLine N x ToString Replace 4 Nr 4 y ToString Replace locEvent locEvent Length 1 Date ToString content AppendLine l content AppendLine global locations locations content AppendLine global events events content AppendLine window File WriteAllText fileName content ToString Therefore the server side module after processeing the request sends the javascript file to the client side module 6 5 2 2 Client side The client side software module was designed with the purpose of sending requests to the server side module and to interpret the answers received As already introduced in section 6 5 2 1 client receives from the server a file with extension js containing the dataset to be processed for the con struction of HeatMap This file is in the following format oO o O OC kt nN
140. net PoE Through the DHCP server it will be then possible to assign the desired IP address to the sensors in order to monitor and send input commands to each individual sensor The sensors must be also connected to each other through the timing Cables CAT5e cables to allow time synchronization among them by calculating the TDOA between each pair of sensors The next step involves the calibration process of the RTLS However you must first perform a virtual reconstruction of the sensors map through the Location Engine Configuration software tool as described in section 5 3 1 2 reference system must be determined within the graphical section of this software tools in order to precisely mapping the sensors The position of each sensor must be manually inserted by bringing the values corresponding to those of the real reference system where the sensors were physically in stalled Therefore it will be possible to carry out the true calibration process which provide the system with the missing parameters in order to perform the localisation The steps to be followed for the calibration of the system are the following sensor orientations and cable offsets In the first phase of calibration also called Orientation calibration once a Ubisense tag is put in a known position within the coverage area of the sensors and the coordinates 5 3 UBISENSE UWB RTLS 47 of the position in the considered reference system are written in the Loca t
141. ng Microsoft R Visual C Express framework NET allowed visualisation of cow s location and speed data In particular with regard to the cow s location data the creation of a CowHeatMap was implemented to allow the visualization of the spatial dis tribution in two dimensions of the data acquired from the RTLS within the study area Two dimensions represent cartesian coordinates x and y values and the third dimension is used for showing the intensity of a datapoint in relative comparison to the absolute maximum of the dataset The different color and color intensity denote the difference in sample density at a location Usually red hot is used for the maxima and blue cold for minima Wied 2011 2013 As regards instead the cow s speed data the building of CowSpeed Graph was implemented to show trend over time of the cow s instantaneous speeds sampled at regular time intervals relatively to data acquired by the RTLS In fact Bell et al 2013 established that deterioration of walking speed is one of the characteristic symptoms of lameness Furthermore mea surement of the speed of each cow showed to be well correlated with the cow s mobility score Bell et al 2013 In addition Martinez Ortiz et al 2013 calculated the speeds of approximately 190 dairy cows inspected by a video tracking system over a number of weeks The relative speed of a cow with respect to the group on its own is not sufficient to detect lamenes
142. nsiderable importance in estrus detection Firk et al 2002 and early diagnosis of lameness Pastell et al 2009 Cow identification has also been frequently obtained by means of Radio Frequency Identification RFID technology which makes use of two main electronic components tags and readers that exchange information through radio waves Schwartzkopf Genswein et al 1999 adopted a RFID system GrowSafe System R for monitoring the feeding patterns of feedlot cattle and demonstrated that RFID technology provides the ability to better under stand feeding patterns thereby aiding in the selection of those management practices that improve animal performance The same system was adopted by Sowell et al 1998 who demonstrated that RFID technology has the potential to identify sick animals and improve feedlot health management practices In another study Voulodimos et al 2010 utilised RFID tags for the storage of information suitable to identify the animal sheep cattle pigs etc in case it gets lost or even collect some basic data e g behaviour against other animals In the same year Samad et al 2010 demonstrated that the use of RFID based identification and data retrieval offers the value added benefit of data security and would prevent insurance related frauds Ilie Zudor et al 2011 carried out a survey of applications and requirements of identification systems and RFID techniques They reported that a new cattle tracking s
143. nt ANALYSIS AND INSTALLATION OF UBISENSE RTLS 63 Figure 6 9 Temperature and humidity sensors anemometers and globe thermometers within the study area A final analysis on the scale of the system was conducted during this pre liminary phase Given the size of the study area it was possible to determine the number of sensors 4 to purchase which were sufficient to cover the entire space involved in the trial and the length of the connection cables to be installed Finally adjacent to the study area and sufficiently isolated from the livestock environment locals were identified where to install the rest of the hardware e g PCs switches and modems 6 2 3 2 Sensors layout During this phase the layout of the sensors through the preliminary definition of their geometric arrangement was designed Sensors Sensso Senswo Sensyg and Senssgg were placed at the four corners of a rectangle outside the area studied which had a size of 15 10 x 13 95 meters at a height of 3 85 meters above the ground as shown in Figure 6 10 The layout of the 4 sensors was not casual since the error committed by the system can also depend on the layout of the sensors as described by Shahi et al 2012 In this work it is recommended that the optimal receiver layout should incorporate spatial variability of the UWB receivers in the three 64 CHAPTER 6 THE CASE STUDY principal directions Thus rectangular or polygonal receiver arrangements are pref
144. number of constraints of speed fixed or variable height max position variance position standard deviation etc re lated to the tags in order to filter measures on the basis of the elements that characterize the usage scenario For our test two filter algorithms CowTag Filter and ReferenceTag Filter were customized in order to associate them to the tags applied to cows and to a tag selected to be installed in a fixed point inside the study area respectively In Figure 6 26 it is possible to observe the set parameters for the two filters in question highlighted by a red box Default fixed height Default information filtering Default no filtering Default static fixed height Default static algorithm ixed height informat information filtering no filtering static fixed height informati static informat stickiness 5 aa gt 5 1 2 0 6 5 0 0 0 0 4 4 NI 0 0 a N Ni PAEZ EN SEPPO OMEO Nen is gt gt Figure 6 26 Filters tab of Location Engine Config and customized filters in the red rectangles One of the differences between two filters for example was the maximum speed of the tags to locate set to 1 m s relative to the Cow Tag Filter since a cow can hardly exceed this speed in these environmental conditions and to 0 m s for the ReferenceTag Filter because the tag was physically installed in a fixed point Further details about all the ot
145. nvolves thousands of animals such as cows pigs laying hens and broilers in highly controlled conditions However nu tritional systems based on products of industrial nature are often adopted and housing systems are organised in functional areas which are not properly managed or provide insufficient space to the animals These conditions are able to increase the risk of causing health problems Therefore the population growth of animals raised in such circumstances leads to a higher probability of creating zoonotic pandemics i e animal infectious diseases that can be transmitted to humans such as avian flu or swine flu The monitoring of animal behaviour should involve the control of growth and production of the herd but also include control of these forms of disease transmitted through direct contact with the skin hair eggs blood or secretions or indirectly through ingestion of contaminated food Another problem which is strictly correlated to the prevention of diseases and the treatment of infections is the use of antibiotics Antibiotics however are used at the same time as a tool to enhance and accelerate growth of the animals This management practice often causes excessive use of these sub stances The direct consequence of this phenomenon is that certain bacteria become resistant to these antibiotics surviving at the exposure to these sub stances T herefore the animals affected by certain diseases could not profit from the th
146. ocal isation error and the related standard deviation In this metric a true positive was assigned when the RTLS detected the tag in the panoramic top view image Therefore the number of true positives was obtained by counting the number of times that the tag was present in the framed scene The number of false positives and the number of false negatives were assumed to be equal to 0 Metric B for each tag all the obtained after the outlier data clean ing process were used to calculate the mean localisation error and the related standard deviation In this metric the number of true positives was obtained by counting the number of times that the tag was present in the framed scene and the related were not filtered out by the data cleaning process The number of false positives and the number of false negatives were assumed to be equal to the number of measurements considered as outliers Metric C a threshold value of 0 50 m was established for localisation error All less than or equal to the threshold were used to calcu late the mean localisation error and the related standard deviation In this metric the number of true positives was obtained by counting the number of times that the RTLS detected the tag with a localisation error lower than the threshold The number of false positives and the number of false negatives were obtained by counting the number of times that the RTLS detected the tag with a localisation error greater 98 CHAP
147. odes In this case the difference between the arrival times of the signals traveling between the target node and the reference nodes is estimated The setup of the resulting localisation principle which is commonly referred to as hyperbolic localisation is shown in Figure 5 8 Figure 5 8 Hyperbolic localisation scheme in TDoA measurements One way to obtain a TDoA measurement is to estimate the ToA at each reference node and then to obtain the difference between the two estimates Specifically if the received signals are given by ri t and r2 t as in 5 6 T is estimated from r t and v is estimated from r2 t Since the target node and the reference nodes are not synchronized the ToA estimates at the reference nodes include a timing offset in addition to the time of flight As the reference nodes are synchronized the timing offset is the same for each ToA estimation Therefore the TDoA measurement can be obtained as TTDoA Ti T2 5 9 where and denote the ToA estimates at the first and second nodes respectively Another way to obtain a TDoA measurement is to perform cross correlations of the received signals r t and r2 t and to calculate the delay correspond ToA measurements provide information about the distance between two nodes by estimating the time of flight of a signal that travels from one node to the other To prevent ambiguity in ToA estimates the two nodes must have a common clock o
148. omogeneous without los ing any relevant information for the analysis performed in this study The data discarded by applying this technique about 1 9 of the error dataset corresponded to measures clearly unreliable Figure 7 12 shows the box plot representation of the error computed on the data provided by the RTLS for each one of the eight tags and the reference tag analysed The measures discarded by the outlier data cleaning were indicated with the symbol In this figure the localisation error at the 25 percentile for the cows tags ranges between 0 m and 0 65 m with an average value of 0 21 m whereas the localisation error at the 75 percentile varies between 0 62 m and 0 97 m with an average value of 0 77 m 136 CHAPTER 7 RESULTS Error m Static Test Points n Point Meanx Meany Meanz 1 0 32 0 26 0 53 277 2 0 25 0 29 0 08 80 3 0 06 0 07 0 05 79 4 0 07 0 0009 0 02 78 5 0 01 0 13 0 21 80 6 0 48 0 36 0 25 21 0 04 0 01 0 24 122 8 0 21 0 02 0 10 41 9 0 18 0 28 0 09 51 10 0 06 0 04 0 07 58 11 0 01 0 16 0 19 41 12 0 09 0 17 0 004 TT 13 0 07 0 11 0 29 76 14 0 07 0 06 0 20 TT 15 0 11 0 05 0 09 79 16 0 04 0 04 0 13 54 17 0 02 0 02 0 001 621 18 0 43 0 04 0 06 79 19 0 15 0 04 0 12 81 20 0 41 0 16 0 42 T7 21 0 23 0 34 0 34 41 22 0 03 0 10 0 27 38 23 0 01 0 006 0 04 39 24 0 04 0 04 0 11 55 25 0 001 0 01 0 01 52 Table 7 1 Planimetric position errors computed on the data provided by the RTLS in the static
149. ortance to avoid delayed inseminations of cows which could reduce cow fertilization rate and as a consequence have negative economic impacts on farm budget and costs The estrus period in fact lasts for 11 16h on average Firk et al 2002 To this aim recent research Brehme et al 2008 J nsson et al 2011 achieved excellent results by adopting real time pedometers for estrus detection in dairy cows However pedometers do not allow cow localisation and cannot be used to distinguish some behavioural patterns e g standing vs feeding perching that are crucial data as it has been proved that the daily incidence of standing behaviour is of considerable importance in estrus detection Firk et al 2002 and early diagnosis of lameness Pastell et al 2009 The software developed in this study offers the ability to work in real time to the user by monitoring the data acquired by the RTLS updated at short time intervals In fact it can be launched during the execution of both the location platform and the tools for recording data in the database and automatically takes constantly updated records from the database as input By adding new control modules this feature would have the advantage of being suitable to notify any inconvenience through alert messages as a result of changes in dairy cow behavior and then instantly alert the farmer Furthermore compared to pedometers that do not allow cow localisation it can be used to distinguish beh
150. p section an example of pan and zoom operation 134 7 11 An example of the computation of the true position of the tags with ID O04 exe UD uu poko ser Sa Bee a wee Ro Sechs x Ee 134 7 12 Box plot of the errors computed on the data provided by the RTLS for each one of the nine tags analysed 138 7 13 screenshot of the software UI at boot time 143 7 14 An example of CowHeatMap displayed in the visualisation section of the BE ux m dk oL e umque p sio dod d ak 145 7 15 An example of CowSpeedGraph displayed in the visualisation section of the DL woke nom EG SE a SS aw 145 7 16 CowSpeedGraph obtained after the zoom operation carried out in Figure TES mox RR RO Rea ce EC 146 7 17 CowHeatMaps and CowSpeedGraphs relative to the tag with ID 020 in CFTime interval distinct by day 1 3 147 7 17 CowHeatMaps and CowSpeedGraphs relative to the tag with ID 020 in CFTime interval distinct by day 2 3 148 7 17 CowHeatMaps and CowSpeedGraphs relative to the tag with ID 020 in CFTime interval distinct by day 3 3 149 7 18 CowHeatMaps and CowSpeedGraphs relative to the tag with ID 020 in CLTime interval distinct by day 1 3 150 7 18 CowHeatMaps and CowSpeedGraphs relative to the tag with ID 020
151. pletely open From an interview with the breeder and the direct observation of the breeding environment knowledge of the management activities carried out in the barn was obtained Feed was delivered to the cows once a day at approximately 06 30 a m and was moved closer to the feed barrier later in the day at 4 30 p m Milking occurred twice a day at 06 00 a m and 05 00 p m Furthermore in the alleys there was presence of slurry accumulation as the cleaning is not automated but it is carried out 1 2 times a day by a scraper Within the barn a reduced but complete breeding area of about 15 40 m x 11 50 m was considered for the experimental trial Figure 6 3 In this area a group of 15 Holstein dairy cows was housed The breeding environment was composed of a resting area of about 10 40 m x 4 30 m with two rows of stalls arranged head to head and sand beds a feeding alley of about 15 40 m x 3 50 m adjacent to the resting area a service passage and two service alleys AREA OF INTEREST 4 1 1 1 1 1 1 ofr 0911 ost U BOX BOX BOX BOX BOX U OFFICE C O Resting area Feeding area Figure 6 2 Plan of the free stall barn showing the area of interest for the experimental trial oric 6 2 ANALYSIS AND INSTALLATION OF THE RTLS 51 o T pic o gt Feeding passage 1 1 1 Manger I a g n i Feeding alley I A A 4 125 Stall
152. quency spectrum allocation of various wireless systems A Global Posi tioning System GPS 1 56 1 61 GHz B Personal Communication System PCS 1 85 1 99 GHz C Microwave ovens cordless phones bluetooth IEEE 802 11b 2 4 2 48 GHz D IEEE 802 11a 5 725 5 825 GHz E UWB 3 1 10 6 GHz The bandwidths and power levels of various systems are not drawn to scale Sahinoglu et al 2008 5 2 LOCALISATION IN ULTRA WIDEBAND SYSTEMS 25 If UWB signals were allowed to transmit over the range of frequencies of these systems without any restrictions all these systems could be jammed by UWB emission Therefore the power spectral density must not overcome 41 3 dBm MHz for frequency ranges from 3 1 to 10 6 GHz and it must be even lower outside this band depending on the specific application In other words the FCC spectral mask specifies a useful spectrum of 7 5 GHz for most UWB systems Sahinoglu et al 2008 5 2 LOCALISATION IN ULTRA WIDEBAND SYSTEMS 5 2 1 Properties of UWB indoor localisation systems As regards the concept of positioning and functioning of the UWB system is necessary to carefully analyze the environment in which localisation is carried out to evaluate some variables that may affect the system and fix the limits of the area to be analyzed in relation to the capacity of the instrument A unit of measure for distance must be established in order to determine the location of a target moving within such an en
153. r by sending UWB signals that are re ceived from the sensors and are used to calculate the Angles Of Arrival AOAs a and a Therefore only two sensors receiving a tag signal are already able to deliver a 3D location 3 Sensors are synchronized via cable Timing Sinchronization via Cable in Figure 5 14 then Time Difference Of Arrival TDOA between each pair of Sensors is calculated This makes the location robust 44 CHAPTER 5 UWB RTLS 4 The Master Sensor calculates a coordinates based on all raw informa tion delivered taking into account tuning parameters such as motion model filtering and sends it to the Location Platform Both TDOA and AOA are used to increase robustness of 3D Position computations Once the location of one or more tags is estimated map window of Lo cation Engine Configuration tool already described in section 5 3 1 2 allows their visualization Among other functionalities the tool provides the ability to view a map of the sensors network in which the AOA vectors are shown in real time through green lines coming from each sensor the hyperbolic lo calisation TDOA with blue hyperbolic curves and the position of the tag with a red dot The Figure 5 15 shows two screenshots of the Location Engine Configu ration tool Figure 5 15 Screenshots del Location Engine Configurator tool for the visualization of 3D e 2D maps respectively Green lines represent the AOA vectors from each sensor Blue hyperbo
154. r they must exchange timing information via certain protocols such as a two way ranging protocol 32 CHAPTER 5 UWB RTLS ing to the largest cross correlation value The cross correlation function can be expressed as Sahinoglu et al 2008 1 T where T is the observation interval and the TDoA estimate is given by Tiol ri t ro dt 5 10 rDoA arg max Tiol 5 11 5 3 UBISENSE ULTRA WIDEBAND REAL TIME LOCATION SYSTEM As already seen in chapter 2 the RTLS used in this thesis Ubisense UK is currently recognised to provide one of the highest accuracies of all the RTLSs Linde 2006 Weichert et al 2010 The achievable accuracy which Ubisense declares for the localisation of moving objects in real time is 15 cm in the three dimensions x y z Steggles amp Gschwind 2005 i e about 12 25 cm in two dimensions However in field tests accuracy of the Ubisense system was found to vary between 30 cm and 100 cm in the two dimensions x and y in dependence on the applications e g agriculture transit yard management and personnel safety Mok et al 2010 Ward 2010 In the next sections the Ubisense RTLS will be analyzed in detail 5 3 1 Platform architecture Ubisense has developed a platform for building Smart Space applications This platform meets the basic requirements for building Smart Spaces ac curate 3D positioning scalable realtime performance development and de ployment tools Accurate 3D loca
155. re physically the same the function of calculating the TDOA between each pair of sensors is performed considering the master as reference sensor for time synchronization For this reason the connection of the Timing Cables must follow a definite scheme The physical scheme of the networks described above is shown in Figure 6 12 while a picture of the connection of the Timing Cables in the master sensor is shown in Figure 6 11 b if cable e PoE cable timing cable DHCP Server Figure 6 12 Scheme of the RTLS hardware used in the trial Both the PC and the switch were placed in a room adjacent to the barn where the two wired networks were converged in order to avoid direct expo ANALYSIS AND INSTALLATION OF UBISENSE RTLS 67 sure to the livestock environment Although not necessary for the functioning of the RTLS also a modem for ADSL connection was connected to the net work in order to allow remote control of the entire system 6 2 3 4 Middleware software installation Completed hardware system operations the software configuration was car ried out Although in Figure 6 12 DHCP Server and PC were represented by two separate physical devices both of them were managed by the same machine To do that a virtual machine where the software provided by Ubisense was installed was configured on the PC while the DHCP server was run on the physical machine The features of the virtual machin
156. rea also in this case by reference to the work conducted by Agosta 2012 in the same period considered in this test and in the same livestock environment CLI 48 in lying 6 8 tot cows In particular CLTime is composed of two time intervals the first in cluded between 00 00 and 05 00 not visible in Figure 6 38 and the second between 11 00 and 14 30 6 5 SOFTWARE TOOL FOR THE VISUAL ANALYSIS 119 000 00 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 Time b Yousel 8GHi cu csi Figure 6 38 CFTime green area and CLTime yellow area defined according to the CFI and the CLI obtained from Agosta 2012 Through the use of the software tools described in this section the data for the selected time period and in the intervals CFTime CLTime were analyzed In particular for each of the seven days in question and for each time in terval established CFTime and CLTime a CowHeatMap and a CowSpeed Graph were obtained In addition the CowSpeedGraphs related to the tag with ID 020 dis tinct by day where the instantaneous velocities calculated by the software in relation to all the 24 hours are represented were analyzed 120 CHAPTER 6 THE CASE STUDY Part IV RESULTS AND DISCUSSION 121 Chapter RESULTS 7 1 LOCALISATION AND IDENTIFICATION PERFORMANCES OF THE RTLS 7 1 1 The RTLS operativity in the barn environment The RTLS showed a good performan
157. registration of the tag to which they were attached The record struc ture remained unchanged In Figure 6 32 the structure of the tags dataset obtained after the execution of such software is shown 90 CHAPTER 6 THE CASE STUDY 15 08 2011 15 08 2011 0 00 02 239 10 1 9 2 0 3 0 01 15 08 2011 0 00 03 239 9 1 8 2 0 9 0 02 15 08 2011 0 00 07 239 10 1 9 5 0 3 0 01 15 08 2011 0 00 10 239 10 5 9 9 0 4 0 03 15 08 2011 0 00 11 239 10 4 9 2 0 3 0 02 15 08 2011 0 00 14 239 10 1 9 4 0 5 0 04 15 08 2011 0 00 15 239 9 9 9 6 0 3 0 01 15 08 2011 0 00 16 239 10 5 9 8 0 4 0 01 15 08 2011 0 00 18 239 10 1 9 9 0 6 0 02 15 08 2011 0 00 19 239 10 3 9 2 0 2 0 02 15 08 2011 0 00 21 239 9 1 9 1 0 3 0 04 15 08 2011 0 00 23 239 10 1 9 0 0 1 0 05 15 08 2011 0 00 27 239 10 8 9 2 0 3 0 01 Figure 6 32 Tag dataset structure Each subset constitutes a folder representing a tag which contains a number of text files equal to the number of days of registration of the tag in question In the figure is also reported an example of the contents of one of these text files corresponding to the day August 15 2011 6 4 1 2 The dataset of the images The tool suitable for recording the images which are captured by the cam eras saved the frames in jpg format in separate folders one for each day by assigning them a unique name corresponding to a number that will be the identifier Therefor
158. resenting a tag which contains a number of text files equal to the number of days of registration of the tag in question In the figure is also reported an example of the contents of one of these text files corresponding to the day August 15 2011 c s ca nosse x ook oo o9 Eoo Y Roos 90 Images dataset structure The number of text files is equal to the number of the days when image recordings were carried out In the figure is also reported an example of the contents of one of these text files correspond ing to the day August 15 2011 i i eoc 2 222 92 Example of computation of the matching between the date of the records related to the dataset of tags and the date of the records related to the dataset of images The arrow shows the matching occurred between records for which one of the first two criteria described above was true 94 Output format after the matching operation 95 Computation of the threshold established for localisation error in Metric 98 Flow chart of the algorithms implemented in the software tool for the visual analysis of the cows location data 100 CF Time green area and CLTime yellow area defined according to the CFI and the CLI obtained from Agosta 2012 119 UI of the tool for RTLS data storage o o c e e sor as e e koso ok aa 124 UI of the software for storing data acquired by the RTLS The displayed tag is applied to a cow in
159. rformance measures in UWB indoor localisation 26 ill iv CONTENTS O0 Positioning se s auos Xv XS X ox aos uos 28 5 2 3 1 Positioning techniques for UWB systems 28 29 UBISENSEUWBILLS 64 mn on Room ESS 32 5 3 1 Platformi 222m oem 32 Solel Hardware architecture os s EX 2 ede as 33 5 3 1 2 Middleware software architecture 38 5 3 1 3 Software applications o 41 5 3 2 Operating specifications 42 53 9 Installation procedure 228850 oo 45 6 THE CASE STUDY 49 61 THE AREA OF THE BARN UNDER STUDY 49 6 2 ANALYSIS AND INSTALLATION OF THE RTLS 51 6 2 1 Requirements analysis 51 6 2 1 1 Functional requirements 51 6 2 1 2 Non functional requirements 52 6 2 1 3 Hardware and software requirements 53 6 2 2 Analysis of the RTLS functionality 53 06221 Use case diagram 2 62 che 54 06222 Class QUUM scio eh be RS SOS 55 0 2 29 Ole diagram ck on mo Re mos 57 06 224 Sequence s y omn 59 6 29 IOS 2s es uos deo So eoo 61 6235 1 SUGBUPVOY ic eo see ee 61 0 2 5 2 Sensors layout RS 63 6 2 8 8 Hardware installation 64 6 2 3 4 Middleware software installation 67 6 2 3 5 System calibration 70 O24 Detergent su Robe SE
160. rors and using precision and sensitivity indices Trade off between these performances was found through the computation of three performance metrics The results showed that in the environmental conditions of the barn the RTLS produced errors which were comparable to the errors declared by the RTLS producer for the fixed reference tag whereas localisation errors related to the tags in movement were higher and in detail a mean error of 0 56 m and an error at the 90th percentile of 1 03 m were obtained Outlier data cleaning produced significant improvements of the results by reducing the average localisation error of about 0 046 m for the cows tags and made data distribution more homogeneous Trade off of RTLS performances yielded an average localisation error equal to about 0 52 m with a position accuracy of 98 for cows tag and an error of about 0 11 m with a nearly 10096 accuracy for the reference tag RTLS performances in the considered environment proved to be generally not dependent on cow behaviour as it is observed for other systems and that RTLS is suitable to determine the occupancy level of the different functional areas of the barn compute cow behavioural indices and track each animal of the herd An automatic and real time software tool for the visual analysis of the cows location data in free stall barns acquired by the RTLS was designed and xvi LIST OF TABLES developed A visual representation used for visualizing two
161. row n 1 l 1 l 1 i 9 11 50 eb 20 3 b I 2 i l 1 1 10 40 tall row n 2 Service alley i l Jg q al 15 40 Plan Figure 6 3 Plan of the study area 6 2 ANALYSIS AND INSTALLATION OF THE UBISENSE ULTRA WIDEBAND REAL TIME LOCATION SYSTEM In this section a detailed requirements analysis section 6 2 1 and the indi vidual steps of the RTLS installation section 6 2 3 will be described 6 2 1 Requirements analysis The requirements for a system are the descriptions of what the system should do the services that it provides and the constraints on its operation These requirements reflect the needs of users for a system that serves a certain purpose such as controlling a device or a peripherical or finding information Sommerville 2010 In this study two kinds of requirements were analysed functional and non functional requirements 6 2 1 1 Functional requirements Functional requirements are statements of services the system should provide how the system should react to particular inputs and how the system should 52 CHAPTER 6 THE CASE STUDY behave in particular situations Sommerville 2010 In this case study the RTLS would have to meet the following require ments Client side View location of sensors on a grid matrix where the map of the barn is reproduced Identify the sensors Activate the UWB tags Calibrate the system View graphi
162. rowband signal time difference can be represented as a phase shift Therefore the combinations of the phase shifted versions of received signals at array elements can be tested for various angles in order to estimate the direction of signal arrival 30 CHAPTER 5 UWB RTLS l 4 er Figure 5 7 Signal arrival at a ULA having a spacing l between the array elements and relation between arrival time differences and AoA In order to obtain theoretical lower bounds on the achievable accuracy of AoA measurements consider a ULA as shown in Figure 5 7 with N antenna elements Let r t denote the received signal at the ith element which is expressed as ri t as t vi nlt 5 6 for i 1 Na where s t is the transmitted signal a is the channel coeff cient t is the delay for the signal arriving at the 7th antenna element and n t is white Gaussian noise with zero mean The delay v can be expressed as d l sin 5 7 with jo Ae i 5 8 for i 1 Na where d is the distance between the transmitter and the center of the antenna array at the receiver and is the inter element spacing Sahinoglu et al 2008 5 2 LOCALISATION IN ULTRA WIDEBAND SYSTEMS 31 Time difference of arrival TDoA TDoA measurements can be obtained even in the absence of synchro nization between the target node and the reference nodes if there is syn chronization among the reference n
163. rtElement curves 14 15 open curve tag with id 1 16 writer WriteStartElement curve 17 writer WriteAttributeString id 1 18 19 open points tag 20 writer WriteStartElement points 21 22 string date 23 for int i 0 i lt instantSpeed2D Count i 24 1 25 open point tag 26 writer WriteStartElement point 27 date instantSpeed2D i Date ToString Replace 28 writer WriteAttributeString x date 29 writer WriteAttributeString y instantSpeed2D i InstantSpeed ToString 30 writer WriteAttributeString ymin 0 31 writer WriteAttributeString ymax instantSpeed2D i InstantSpeed ToString 32 close point tag 33 writer WriteEndElement 34 35 36 37 38 39 40 Al 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 110 CHAPTER 6 THE CASE STUDY close points tag writer WriteEndElement open and close label tag writer WriteElementString label SPEED close curve tag writer WriteEndElement close curves tag writer WriteEndElement open and close description tag writer WriteElementString description elt Speed Graph amp 1lt close plot tag writer WriteEndElement writer WriteEndDocument writer Close The file obtained with the instructions just shown is in following format lt xml ver
164. ry cows Computers and Electronics in Agriculture 72 119 126 Tullo E Fontana I amp Guarino M 2013 Precision livestock farming an overview of image and sound labelling In Precision livestock farming 2013 p 30 38 Leuven Belgium Ubisense 2008 2010 Ubisense hot to articles Retrieved June 2013 from http eval ubisense net howto Contents Comtents Htmi Ubisense 2013a Ubisense location platform Real time location plat form middleware and tools Retrieved June 2013 from http www ubisense net en media pdfs factsheets pdf 81357 ubisense location platform pdf Ubisense 2013b Ubisense series 7000 compact tag fact sheet rev 5 en120812 Retrieved november 2013 from www ubisense net en resources Voulodimos A Patrikakis C Sideridis A Ntafis V amp Xylouri E 2010 A complete farm management system based on animal identification using rfid technology Computers and Electronics in Agriculture 70 380 388 Ward A 2010 Ultrawideband in building location systems Retrieved June 2013 from http nrc simtech a star edu sg rfid slot downloads d401 deaf9251c_u1762 pdf Wathes C M 2010 The prospects for precision livestock farming Journal of the Royal Agricultural Society of England 171 26 32 Weichert F Fiedler D Hegenberg J M ller H Prasse C Roidl M et al 2010 Marker based tracking in support of rfid controlled material flow systems Logistics Research 2 13
165. s Finkenzeller 2003 Ruiz Garcia amp Lunadei 2011 In order to avoid introducing foreign objects into the animal body the small size and weight of the tags made it possible to fix them in the animals ears by means of a suitable plier as shown in Figure 6 23 a To ensure resistance to water mud humidity and impact tags tags were covered with adhesive insulating material and placed in a waterproof bag constituted by synthetic leather and plastic material as shown in Figure 6 23 b ANALYSIS AND INSTALLATION OF UBISENSE RTLS 81 a Figure 6 23 b Protecting wrapping of tags front and back and a Compact Ubisense Tag a Application of the tags to three cows However referring to another study where the tag application method regarded tag positioning on cow s collar and on special bands Huhtala et al 2007 the use of an additional technique of tag application was assessed in order to avoid infections or modifications in cow s behavior due to the weight on the ear The application of the tags on cow s collar is part of an other trial of this research which will be described in section 6 5 4 By using plastic cable ties eight Compact Ubisense tags were applied to eight collars and subsequently covered by adhesive tape to isolate them from water mud and moisture In Figure 6 24 the result is shown a a tag installed in a collar b the outer side of the collar c the inner side of the collar d the eight collars
166. s a cow may be consistently slower than the group due to old age or simply due to its own preferred pace of walking Therefore they proposed to detect lameness by monitoring each individual cow s speed over a time interval to look for consistent changes in mean speed On the basis of this considerations the 100 CHAPTER 6 THE CASE STUDY CowSpeedGraph function makes it possible to monitor each individual cow s speed over a time interval selected by the user This software described in detail in the next sections sections 6 5 1 sections 6 5 2 and sections 6 5 3 was designed to have the necessary features to be integrated into the world of IoT giving to the user the ability to work in real time by monitoring the data acquired by the RTLS updated at short intervals of time In fact the software can be launched during the execution of the location platform and the tool for recording data in the database and automatically takes records of the constantly updated database as input By adding new control modules this feature has the ability to raise alert messages from changes in dairy cow behaviour and then to alert immediately the farmer when some health problems e g lameness or a particular physiological status e g estrus occur In Figure 6 37 the flow chart of the algorithms implemented in this soft ware is shown Server database Vortag Pa eL rss Data selection by f IdTag and Date Y CowHeatMap Implementation Bi E ed
167. s and military sites where high reliability and low latency are essential The Location Platform includes a suite of software tools for easy system configuration and its open API covers every service feature so that anything that can be done using a tool can also be done via the API Location Engine Configuration The Location Engine Configuration tool is used to control and monitor operation of location sensors and tags as well as to tune the location system behaviour The tool enables Monitoring and control of the sensors that are running across the plat form Monitoring and control of the tags that have been registered to the platform 5 3 UBISENSE UWB RTLS 39 Defining the cell functionality e g conventional radio settings and cell boundaries Defining the sensor functionality e g master slave operation and thresh olds Sensor calibration Configuration of the filter mechanisms for the location calculation in the master sensors Processing presence information from a combination of complementary indoor outdoor positioning technologies Map The map tool allows visualization of objects tracked using the Ubisense system It includes two main elements 2D Viewing to view objects in a map like plan view 3D Viewing to view objects in a perspective or virtual reality view Platform Control Platform Control is used to control the core services that make up the Ubisense Location Platform The tool interacts with
168. showed that a higher error is achieved than that of the reference tag However this error keeps under 1 m which is the threshold that Mok et al 2010 considered achievable by the system under adverse geometrical site conditions As regards the localisation error of fixed tags in dairy houses T gersen et al 2010 obtained a precision of approximately 0 6 m by using the Bluetooth wireless technology Huhtala et al 2007 obtained a precision of 1 m for the 7096 of the measures and 2 m for the 90 of the measures by using a WLAN Wireless Local Area Network Very poor results were obtained Huhtala et al 2007 for tags in motion since errors were higher than 3 m Another feasibility study Zhou amp Shi 2009 of localisation with passive RFID tags using multilateration and Bayesian inference algorithms which was carried out in an open field obtained mean localisation errors of 0 19 m with a standard deviation of 0 24 m and 0 37 m with a standard deviation of 0 11 m for the two algorithms respectively The comparison between these research outcomes and those obtained by the RTLS utilised in this study shows that the localisation performance of the RTLS based on UWB technology for tags in movement is comparable to that of other localisation systems for fixed tags Since the worst mean localisation error tag ID 023 which was about 0 80 m is small if compared with the average dimensions of a cow the achieved results led to affirm
169. sion 1 0 gt plot id 2 gt lt curves gt lt curve id 1 gt lt points gt N O0 OC A Q 10 11 12 13 14 15 16 17 18 point x 17 04 2013 6 16 03 y 0 ymin 0 ymax Q gt point x 17 04 2013 6216 08 y4 0 0123 ymmin 0 ymax 0 0123 gt lt point x 17 04 2013 6 16 13 y 0 0083 ymin 0 ymax 0 0083 gt point x 177 04 2013 62 16 18 y 0 0297 ymin 0 ymax 0 0297 gt poinmt x 17 04 2013 9315243 0119 ymin 0 ymax 0 0119 gt point x 17 04 2013 92 15 48 y 0 043 ymin 0 ymax 0 043 gt point x 17 04 2013 9 15 53 y 0 0441 ymin 0 ymax 0 0441 gt lt points gt 19 21 22 23 6 5 SOFTWARE TOOL FOR THE VISUAL ANALYSIS 111 lt label gt SPEED lt label gt lt curve gt lt curves gt description gt lt CDATA Cow Speed Graph gt lt description gt lt plot gt A plot is defined by an identifier id 2 in this example zero or multiple curves id 1 in this example and a points list of points made up of objects of point type Each of these objects is composed by the couple x y to be plotted where x corresponds to the sampling instant and y corresponding to the value of the instantaneous velocity There are also the ymin ymaz values used to indicate the minimum point and the maximum point respectively of y axis in order to highlight a specific part o
170. st ensure a trade off be tween the extensibility of coverage in cases in which you want to add additional space to perform the localisation in neighboring areas and the capacity to be able to locate a growing number of the mobile target simultaneously Robustness This parameter for evaluating the performance of a lo calisation system expresses the ability of a system to be immune to 28 CHAPTER 5 UWB RTLS interference with other systems but also to the variations of the envi ronmental features such as temperature and humidity Device size and weight Especially in cases where we want to locate animals or humans it is important to take into consideration the size and weight of the transmitter devices active tags that need to be transported If these are bulky and heavy the natural behavior of users could be modified Linde 2006 5 2 3 Positioning UWB systems thanks to the brevity of the pulse that characterizes them are ideal candidates for combining comunication and positioning at the same time The duration of a pulse is inversely proportional to the bandwidth of the transmitted signal If the arrival time of a pulse is known with little uncertainty then it is possible to accurately estimate the distance covered by the impulse from the source Combining the distances estimated by multiple receivers we can use simple triangulation techniques to estimate the position of the source For UWB systems with a bandwidth of 7 5 G
171. stalled To perform this operation follow the standard procedure of including a library as shown in Figure 5 12 and in Figure 5 13 Solution Explorer Solution SimpleUbisenseApp 1p A X od Solution SimpleUbisenseApp 1 project c SimpleUbisenseApp W Properties Add Reference B Mr Add Web Reference System Drawing 43 System Windows Forms C3 System xml H EE Formt cs c Program cs Figure 5 12 Add references in a Microsoft R Visual Studio Express project Step 1 first click with the right mouse button on the References of a project and then the left mouse button on Add Reference 42 CHAPTER 5 UWB RTLS Add Reference COM Projects Browse Recent Look in 9 bin Q u ubisense tunnel Forwarder exe Jubisense_undeploy exe UbisensePlatformControl exe JubisenseDataDictionary exe UbisenseRemoteAssistance exe A UbisenseLocationEngine dll UbisenseSaveConfiguration exe UbisenseLocationEngineConfig exe ROT UbisenseServiceInstaller exe UbisenseServiceManager exe UbisenseSimulator exe UbisenseMap exe UbisenseMulticastAdmin exe Filename bisensePlatfarm dl UbisenseLocationServices dll Files of type Component Files dll tlb olb ocx manifest v JL Figure 5 13 Add references in a Microsoft R Visual Studio Express project Step 2 select the desired
172. stored recording of the two systems for which you want to perform the visual recognition of the position of each tag Once that the date is selected the software loads and displays the in formation captured by the two systems in relation to the first record in the input file corresponding to the selected day The software UI is composed of different sections which are utilised to display some information or interact with it by sending some commands Figure 7 6 7 1 RTLS PERFORMANCES 129 University of Catania Department of Agri food and Environmental Systems Management Les mm Day Select 2011 08 23 Control Goto Cusrert Postion 8010 26720 lt 30 e 1 39 TAG 08 0 TrackBar Position 8010 Image Date 23 8 2011 10 31 1 Curent Tot Tag Tape a TAG 87 x Y z TAG 26 x Y z VM 395 10416 1874 ZE 3516 8543 0574 Std Std 8 0 00055963802151382 0 000994095928035676 Date 23 08 2011 10 31 01 23 08 2011 10 31 01 TAG 254 TAG 04 X Y 2 X z E Y 477 735 1403 4785 4661 0624 8 0000297435792163014 Std Bd 01000475362641736865 NIOUNTONYIN OTSTINVIDOVWO HO VISIO Date 23 08 2011 10 31 01 Date 23 08 2011 10 31 02 t VERE Y z hoi Y z 9692 104 V 6826 1211 Sd S m 2 1851319615962E06 0 00215845741331577 Date 23 08 2011 10 31 02 Date 23 08 2011 10 31 01 TAG 239 TAG 20
173. t record and the reproduction of the plan view of the study area is available for the user The software allows the visualization of current tags positions within the Map by using graphic elements points above which the last three digits of the respective tag are shown The software also provides the user with nine labels each corresponding to one of the tags applied to cows These labels are located just above the Map as it is shown in Figure 7 8 1110311 755010 Figure 7 8 The nine labels each corresponding to one of the tags applied to cows By using these labels the user has the possibility to accurately determine the true position of each tag by placing them appropriately to the desired position by a drag and drop operation Through a visual recognition of the Map and the use of these labels the user has the possibility to accurately determine the true position of each tag by appropriately placing these labels to the desired position through a drag and drop operation During this operation on the top right corner of the UI the coordinates x y of the true positions of the tags are shown During this phase the identikit tab is an important support tool for the user because it makes it possible to check whether the tag was successfully associated to the cow in question by the RTLS In this section there are three buttons Apply it makes it possible to apply the changes without saving on disk
174. tag calibration in order to calculate the AOA between the tag and the sensor The blue curves indicate the TDOA between each pair of sensors Ubisense 2008 2010 Once the installation procedure is completed you can use the RTLS for 48 CHAPTER 5 UWB RTLS the localisation and tracking of tags within the area covered by the sensors Chapter 6 THE CASE STUDY The trial was carried out from 1st August to 10th September 2011 and from 1st June to 30th September 2013 within a dairy house located in the province of Ragusa Sicily Italy which is one of the most important livestock breed ing areas in the country Figure 6 1 Figure 6 1 Province of Ragusa Sicily Italy one of the most important livestock breed ing areas in the country 61 THEAREA OF THE BARN UNDER STUDY The barn was characterized by a rectangular plan of about 55 6 m x 20 7 m with three sides completely open i e without outside walls Figure 6 2 The roof was symmetric and covered by fibro cement sheets supported by 49 50 CHAPTER 6 THE CASE STUDY a bearing structure made of steel trusses and purlins The feeding alley of about 55 75 m x 3 50 m was adjacent to the resting area that was arranged with two rows of 64 stalls faced head to head equipped with sand beds Ser vice alleys allowed the easy access of the cows from the feeding alley to the service alley for the second row of stalls The side of the barn at the back of the second row of stalls was com
175. tallation Procedure of Ubisense system structured in five steps Ward PU CC ee a ee 45 Map of sensor network after calibration operations Each sensor is ori ented toward the calibration point the center where the tag used for calibration red dot is placed The green lines outgoing from each sen sor focus the tag calibration in order to calculate the AOA between the tag and the sensor The blue curves indicate the TDOA between each pair of sensors Ubisense 2008 2010 4T Province of Ragusa Sicily Italy one of the most important livestock breeding areas in the 49 Plan of the free stall barn showing the area of interest for the experimen EME eT uoce Bik Ron Red Rh Eee x mon d uestes 50 Plan of the study area 2 2 22 51 Use ease diarani oo ecs os kee ee Bene Gee eee 55 Class i re a A 9 x39 RR 56 Shate doe dono so dex Se x Wow m ho YOUR 58 Sequence diagrama o 4 Hee 60 Photo of the study area 62 Temperature and humidity sensors anemometers and globe thermome ters within the study area c eos ea eres 63 LIST OF FIGURES 6 10 6 11 6 12 6 13 6 14 6 15 6 16 6 17 6 18 6 19 6 20 6 21 6 22 6 23 6 24 6 25 6 26 6 27 6 28 6 2
176. test 7 1 RTLS PERFORMANCES Tag ID 004 008 020 023 026 053 239 254 187 Error m before outlier data cleaning Min Mean Max 90 perc 0 12 0 59 4 41 1 12 0 15 042 2 96 0 98 0 17 0 42 4 16 0 87 0 18 0 83 2 62 1 20 0 16 O73 6 84 1 25 0 09 049 3 97 0 90 0 14 045 3 97 0 96 0 17 0 56 5 81 0 97 0 00 0 11 0 56 0 17 Points n 1029 1582 1377 1453 1358 1362 1209 1372 3672 137 Table 7 2 Planimetric position errors computed on the data provided by the RTLS for each one of the eight tags analysed and for the reference tag ID 187 before outlier data cleaning Tag ID 004 008 020 023 026 053 239 254 187 after outlier data cleaning Min Mean Max 90 perc 0 12 0 53 1 52 1 04 0 15 0 39 1 66 0 93 0 17 0 38 1 56 0 82 0 18 0 79 1 44 1 12 0 16 0 66 1 71 1 17 0 09 0 45 1 68 0 87 0 14 0 41 1 87 0 91 0 17 0 51 1 33 0 88 0 00 011 0 24 0 17 Points n 1029 1582 1377 1453 1358 1362 1209 1872 3672 Table 7 3 Planimetric position errors computed on the data provided by the RTLS for each one of the eight tags analysed and for the reference tag ID 187 after outlier data cleaning 138 CHAPTER 7 RESULTS Error Figure 7 12 Box plot of the errors computed on the data provided by the RTLS for each one of the nine tags analysed After the outlier data cleaning technique was applied to the dataset the average values of the mean error and the
177. that the RTLS could be used for studying some spe cific aspects of cow behaviour For instance this error would not affect the computation of some behavioural indices that do not require a high level of precision on the position of the cow such as cow standing index and cow feeding index Bava et al 2012 Mattachini et al 2011 Overton et al 2002 Provolo amp Riva 2009 Moreover a relatively small localisation error shows that the RTLS has the capability to give good description of the occupancy level of the different functional areas of the barn as well as that it could be utilised to track each animal of the herd with a good approximation Further improvements may contribute to the reduction of the mean lo calisation error through the application of predictive algorithms such as the Kalman filter which deal with the problem to reduce the amount of noise present in a signal by comparison with an estimation of the desired noiseless signal Kalman 1960 From visual analysis of the video recordings it was observed that differ ent cow behaviours corresponded to similar values of mean error and error distribution For instance the tag having ID 008 which corresponded to a cow standing still in the area during the time interval considered showed a mean error equal to that of the tag having ID 020 which was associated to 157 a cow in movement and a similar error distribution Figure 5 Therefore this would lead to affirm that R
178. the material of which they are made e g if there are cement walls metal doors 26 CHAPTER 5 UWB RTLS stairs beams etc in order to determine the possible degree of reflection scattering and attenuation they produce in the presence of electromagnetic waves Furthermore a study on the level of air temperature and relative humidity which should be suitably measured is recommended in order to determine the parameters necessary for the choice of the location technology to be adopted 5 2 2 Performance measures in UWB indoor localisa tion A localisation system should fulfill certain features relating to its performance in order to be considered suitable for their intended use The following parameters are the main performance measures Accuracy Measurement accuracy reflects the closeness between the measurement result and the true value of the measurand Accuracy is a positive characteristic of the measurement but in reality it is ex pressed through a dual negative characteristic inaccuracy of the measurement The inaccuracy reflects the unavoidable imperfection of a measurement The inaccuracy of a measurement is expressed as the deviation of the measurement result from the true value of the mea surand this value is called measurement error or as an interval that covers the true value of the measurand Rabinovich 2010 Precision This parameter indicates the degree of reproducibility of a measurement Typically the ac
179. the core platform com ponent of the Ubisense Domain Object Model This core component consists of The platform itself which coordinates between controller nodes on the sensor network Controller nodes which provide the operating environment for the var ious services running on each machine Security Manager Ubisense provides security and access control through the Security Man ager The Security Manager allows the definition of a security policy for the Ubisense Location Platform to protect platform data from unauthorized ac cess It is designed to ensure that only authorized users are able to modify the Ubisense Location Platform state 40 CHAPTER 5 UWB RTLS Service Installer The Service Installer is a tool that allows extra services or service upgrades to be installed into the Ubisense Location Platform intuitive Service Installation Wizard guides the user through the installation process Platform services are installed from a user defined distribution directory Service Manager The Service Manager is used to administer the distributed services that make up a Ubisense platform The tool supports Monitoring the services that are running across the platform Starting and stopping services Deploying services on service controllers Requesting that services backup their data to the core server Site Manager The Ubisense Site Manager simplifies the configuration of the properties of types objects and building
180. time liness of results production The system considering its scope can be called a soft real time system as it can tolerate an occasional viola tion of a deadline causing in the worst case a irrelevant degradation of performance but not the failure Oshana 2006 In fact regardless of the chosen deadline value the goal for which the system was designed does not require response times in the order of milliseconds Reliability when using the system no serious malfunctions were ob served Robustness the system behaved in a reasonable manner in unex pected situations which are not covered by the specifications An ex ample of an unexpected situation was the presence of certain records in the dataset containing the value of in the instantaneous speed field l The results of the carried out processing in response to a certain event must be pro duced within a given time interval said deadline 158 CHAPTER 8 DISCUSSION due to control which was not originally planned during the program ming phase The software computed the output correctly discarding however this exception only in the case of CowHeatMap while the CowSpeedGraph reported errors The bug however was subsequently solved E ciency resources memory CPU etc were used in a proportion ate way for the services that the software performed i e no waste of resources was encoured Usability The UI is very simple and user friendly and does not require
181. to the cartesian reference system determined in the design phase of the sen sors layout section 6 2 3 2 the coordinates in three dimensional Euclidean space of these points were measured Next in the considered area a tag was introduced and at intervals of two minutes it was placed in each of the selected points allowing the RTLS to capture and record its position at 2 seconds intervals Obviously the measurements acquired by the RTLS in the instants where the tag was moved from one point to another were discarded Figure 6 22 shows the layout of the points chosen for the realization of this test 80 CHAPTER 6 THE CASE STUDY o MOM EE Sens Sens tia d 15 10 1 x Feeding passage l j Manger Tag rif i o Feeding alley al row n 95 9 9 o e a gt E 5 a 3 c 4 E e Seto 2 S Service alley E ens o E o n iT E Static test point xX Sensor IP30 Calibration point 1 2 3 4m Figure 6 22 Static test points and calibration point in the area of the barn under study 6 2 7 Cow tags application Once the static test was carried out eight active tags were installed on the eight cows In literature different techniques was used to apply RFID tags to the body of the animal
182. ttons The Compact tag provides a button to allow context sensitive input in systems that require interactivity Applica tions can use the location of the tag when the button is pressed and perform functions based on an event for example the activation of a production process of a machine when we press the button but only if the user is located in a safe place The application can also send feedback to the user via the LED Resistant and adaptable The tags are designed to withstand to harsh industrial environments They resist to dust water moisture and impacts and can be safely installed on mechanical or electronic instrumentation and animals as well as humans Battery life The techniques of low energy or power management affect the battery life In a typical application where a tag is used to identify a user every three seconds the battery has an average life of four years Charge level of the battery is shown by the system to allow a planned replacement the actual duration will depend upon the type of battery tag and blink rate Ubisense 2013b Optional additional receivers and sensors The Ubisense platform allows the integration of additional devices such as measuring systems transducers sensors detectors e g GPS 802 11 RFID 38 CHAPTER 5 UWB RTLS temperature sensors radiometers This allows a usage of a variety of equip ment to meet the needs of users while avoiding blocking the user to a single vendor
183. tware Initially database updating was automated in order to allow the backup of data in such data structure By using a timer a time is set which the follows operations are cyclically iterated e Getting all object out of Ubisense database e Event handler of Ubisense Location Update e Storing cow location data in a temporary data structure The first two operations as described in the previous section implement the functions of reading the records in the Ubisense database and updating the variables where the extracted data are stored respectively Subsequently this information was stored in a temporary data structure 78 CHAPTER 6 THE CASE STUDY Setting update time of the database with a timer Getting all object out of Ubisense Database Event Handler of Ubisense Location Update Storing cow location data in a temporary data structure Timer expired yes Connect to local SQL Server database y SQL query create table with cow location data Cow location SQL query Insert cow location data In table and then delete cow location data from temporary data structure Disconnect from local SQL Server database Figure 6 21 Flow chart of the software for the automatic storage of data acquired by the RTLS on a database ANALYSIS AND INSTALLATION OF UBISENSE RTLS 79 When the time set in the timer expires the software starts a connection to a SQL Server database previously installed
184. ue index j else if sought lt v m date j m 1 else i m 1 while i lt j amp amp found false return found 4 called Matching tagName yyyyl mm dd txt 94 CHAPTER 6 THE CASE STUDY In Figure 6 34 is shown an example of computation of the matching be tween the date of the records related to the dataset of tags and the date of the records related to the dataset of images The arrow shows the matching occurred between records for which one of the first two criteria described above was true ID IMAGE DATE IMAGE ID TAG DATE TAG HH MM SS HH MM SS 242903 06 01 44 51 06 01 44 242906 06 01 50 55 06 01 48 242910 06 01 58 57 06 01 49 242911 06 01 59 58 06 01 50 242912 06 02 02 m 65 06 01 56 242913 06 02 04 SS 68 06 01 59 242914 06 02 06 T m 06 02 07 242915 06 02 08 79 06 02 08 242916 06 02 10 80 06 02 09 242917 06 02 11 81 06 02 10 242921 06 02 20 83 06 02 12 242922 06 02 22 84 06 02 13 242923 06 02 24 85 06 02 14 242924 06 02 26 R 99 06 02 26 242925 06 02 28 101 06 02 27 103 06 02 28 Figure 6 34 Example of computation of the matching between the date of the records related to the dataset of tags and the date of the records related to the dataset of images The arrow shows the matching occurred between records
185. uestion was in perching The software is able to clarify the cow behaviour observed because it shows the graphic element corresponding to the tag located in a cubicle and at the same time the value of the z coordinate equal to 1 256 m which are typical features of the behaviour associated with perching ES University of Catania Department of Agri food and Environmental Systems Management PEGIEC Fie Total 10 020000086004 Current 020000086020 T 3 020000085020 020000086023 020000086026 Update 20 04 2013 12 11 08 620 020000121239 tme 020000121254 020000139187 x Q200001 Current X 10 083 1 p Cunent Y 0 0 19142 2 1 256 Std Enor 2 5387396812439 Hide Map Figure 7 2 UI of the software for storing data acquired by the RTLS The displayed tag is applied to a cow in perching Figure 7 3 instead shows the data relating to one of the cows tags in the case where the cow in question was in Lying The software in fact shows the graphic element corresponding to the monitored tag within a cubicle and simultaneously the value of the z coordinate equal to 0 345 m which are typical features of the behavior associated with the Lying 126 CHAPTER 7 RESULTS Fie wa RE 020000086004 020000139187 020000086008 tag Id 020000086020 020000086023 020000086026 Update 20 04 2013 12 09 20 865 020000121239 e EUH _ 123187 Current X
186. utomatic storage of data acquired by the RTLS on a database lt sa cor s ao ed oom o on oe o Ron o Static test points and calibration point in the area of the barn under study b Protecting wrapping of tags front and back and a Compact Ubisense Tag a Application of the tags to three Preparation of the collars a a tag installed in a collar using plastic cable ties b the outer side of the collar c the inner side of the collar d the eight collars after having isolated the tags with an adhesive tape Application of a collar equipped with tag toa cow a Filters tab of Location Engine Config and customized filters in the red cernucj s e ak Gi Be PO 2D view of the object of the test in the LEC tool 3D view of the study area The displayed tag is one of the tags applied tO Phe COWS 2o xocox dedo x ow dos 8o deos cx uno E debes Seb 3D view of the study area The displayed tag is the reference tag Another point of view of the reference tag in Hardware scheme of the video recording system ix 64 65 66 68 68 69 69 71 12 76 TT 78 80 8l 8l 82 83 84 6 32 6 33 6 34 6 35 6 36 6 37 6 38 P T2 7 3 7 4 7 5 7 6 1 7 7 8 LIST OF FIGURES Tag dataset structure Each subset constitutes a folder rep
187. vious study Barbari et al 2008 could be solved by using the type of tag utilised in this trial Since the application of tags in cow s ears could cause infections or modifications in cow s behaviour due to the weight on the ear further improvement of the tag application methods regarded tag positioning on cow s collar In fact the application of the tags on cow s collar is a totally non invasive method offers the possibility to integrate additional devices for other studies guarantees a longer duration in time allow for an easier intervention in case of maintenance of tags due to the ease of insertion removal of the collars Furthermore the collars are accessories easily commercially available and do not require high costs Although the outlier data cleaning was carried out for the eight tags and the reference tag the results related to the reference tag were not included in the computation of the because this tag has lower errors due to its fixed position and therefore this would lower the mean error of the tags as a whole In the operative conditions considered in this research the results proved that the RTLS produced an error in computing reference tag positions which is comparable with that declared by Ubisense and smaller than that obtained by Stephan et al 2009 under optimal operating conditions and under re 155 156 CHAPTER 8 DISCUSSION alistic shop floor conditions The results related to the tags applied to the cows
188. vironment Furthermore a two dimensional or three dimensional reference system must be defined in the environment in order to locate the target For instance a system of Cartesian or polar coordinates can be used in order to describe the x y and z coordinates or latitude longitude and altitude respectively In some cases it may be more appropriate to adopt a location logical model For example in a context in which we want to determine the presence or absence of a cow in a particular functional area within a livestock environment we can use a labeling system feeding alley service alley bunks etc to refer to such areas This logic model can be derived from a location geometric model The scale of the localisation system must not exceed the threshold of a few hundred meters The localisation systems adopted in a restricted area are usually referred to as Local Positioning Systems LPS unlike to those used on a large scale such as 2G 3G cellular network positioning systems or the Global Navigation Satellite Systems GNSS The indoor location systems fall into the category LPS In general the positioning can be performed using relative or absolute localisation methods or even by making use of both methods Linde 2006 A further analysis of the environment concerns the study of the physical features i e environmental conditions where we want to implement the lo calisation system It is important to take into account the obstacles and
189. vity in the barn environment 123 7 1 2 The software for storage of the data acquired by the POU Cnr 123 7 1 3 The sequence of panoramic top view images and the building of the dataset of tags true positions 126 7 1 3 1 User Interface of the software for the compu tation of tags true positions dataset 127 7 1 4 Planimetric position errors obtained by the RTLS 135 7 1 5 Trade off between localisation and identification per formances of the RTLS 139 7 2 SOFTWARE TOOL FOR THE VISUAL ANALYSIS 141 7 2 1 Features of the software 141 Pall CowHeatMap 2 5 64 4 eo 6 90 ES 142 i12 CowDDeegGISUB s ev ed eee Seem da 142 7 2 2 Software User Interface s sod ea 2244 fee ee oe 143 7 2 3 Use case of the software tool 146 8 DISCUSSION 155 vi CONTENTS V CONCLUSIONS 161 9 Conclusions 163 References 165 List of Figures 5 1 5 2 5 3 5 4 5 9 5 6 9 7 5 8 5 9 Frequency spectrum and power level of UWB signal in comparison to other wireless technologies Domain 20013 Characteristics of a UWB signal absolute bandwidth B is at least 500 MHz or fractional bandwidth byrac fe is larger than 0 2 Gezici et 81 2005 2 ace tie eee mon hE ee OEE Heme E Red The UWB pulse gaussian doublet according to Equation 5 5 with 0 4 ns and width of around 1 ns Sahinoglu Gezici amp Guvenc 2008
190. where the instantaneous velocities calculated by the software in relation to all the 24 hours are represented By comparing these charts is even more evident that the curve of the instantaneous velocity on 12th August 2013 shows higher values compared to those recorded in other days especially in the time interval ranging from the 08 00 am to the 11 30 am 7 2 SOFTWARE TOOL FOR THE VISUAL ANALYSIS 0200 0600 0800 1000 1200 1400 1600 1600 2000 2240 Time a CowSpeedGraph Tag Id 020 Date 10 08 2013 Interval 24h os 2013 08 12 0000 55 COW SPEED 0 06 BU TN 14 00 22 00 07 QE 060 000 1600 1800 X40 c CowSpeedGraph Tag Id 020 Date 12 08 2013 Interval 24h 08 2013 08 14 00 00 02 COW SPEED 0 06 05 04 03 Cow s Speed m s 02 lalia la i al RR i he ha ali POON 02 0 04 00 0600 0800 1000 1200 1400 1600 1800 2000 22 00 Time e CowSpeedGraph Tag Id 020 Date 14 08 2013 Interval 24h 1 P 0230 0400 0600 08 00 1000 1200 1400 1500 18 00 2090 22 00 Time 2013 08 1 1 00 00 02 COW SPEED 0 0240 0400 0600 0800 1000 1200 1400 1600 1890 2040 2200 Time b CowSpeedGraph Tag I
191. which allows the interfacing of RTLS with Corporate Systems and the Applications level that contains the software applications customized to the use case in question Heathcote 2011 5 3 1 1 Hardware architecture The hardware architecture of the Ubisense system consists mainly of sensors tags and optional additional receivers and sensors Sensors Ubisense sensors series 7000 Figure 5 10 Table are tools for precise measurements based on an array of 5 UWB antennas to receive radio pulses 34 CHAPTER 5 UWB RTLS Figure 5 10 Ubisense Series 7000 IP30 Sensors The sensors calculate the position of tags on the bases of the reception of Ubisense UWB signals pulses transmitted by the tag Each sensor independently determines via the antenna array both the azimuth and elevation AoA of the UWB signal obtaining a vector for each tag The TDOA is determined between pairs or more of sensors connected to each other via a synchronization cable This combination of measurement techniques AoA and TDOA provides a flexible tool for a powerful and robust localisation system which allows both an accurate 2D position de termined by a single sensor and a precise 3D position determined by two or more sensors receiving the UWB signal pulse This reduces infrastructure requirements reducing costs while achieving high reliability and robustness of the system Belowe there are some typical features of Ubisense sensors Reactivity
192. with reference to each metric precision and sensitivity are shown for identification performance and mean error and standard deviation are shown for localisation performance Metric A Tag Identification Localisation ID Mean Error Precision Sensitivity Std m 004 1 1 0 59 0 49 008 1 1 0 42 0 47 020 1 1 0 42 0 44 023 1 1 0 83 0 34 026 1 1 0 73 0 59 053 1 1 0 49 0 46 239 1 1 0 45 0 49 254 1 1 0 56 0 44 187 1 1 0 11 0 05 Table 7 4 RTLS identification and localisation performances computed for each one of the eight tags analysed and for the reference tag ID 187 for the Metric A The average values of precision sensitivity and mean error standard deviation of the eight tags resulted about equal to the following values re spectively Metric A 1 1 0 56 0 46 Metric B 0 98 0 98 0 52 0 36 Metric C 0 47 0 47 0 18 0 20 From analysis of these values a high improvement of the localisation performance was observed for Metric C when compared with Metric and Metric B On the contrary identification performance worsened from Metric to Metric B and particularly in Metric C The comparison between the 140 CHAPTER 7 RESULTS Metric B Tag Identification Localisation ID Mean Error Precision Sensitivity Std m 004 0 97 0 97 0 53 0 38 008 0 98 0 98 0 39 0 41 020 0 98 0 98 0 38 0 34 023 0 96 0 96 0 79 0 27 026 0 97 0 97 0 66 0 39 053 0 98 0 98 0 45 0 35 239 0 98 0 98 0 41 0 40 254 0 96 0 96 0 51 0
193. y splitDatel 0 month splitDatel 1 year splitDatel 2 date year month day 4 time content AppendLine new Date date pep ur instantSpeed2D i InstantSpeed ToString Replace date instantSpeed2D instantSpeed2D Count 1 Date ToString Replace splitDate date Split time splitDate 1 splitDatel splitDate 0 Split day splitDatel 0 month splitDatel 1 year splitDatel 2 date year month day time content AppendLine new Date date instantSpeed2D instantSpeed2D Count 1 InstantSpeed ToString Replace m content AppendLine pays content AppendLine content AppendLine labels Date COW SPEED content AppendLine fillGraph true content AppendLine ems content AppendLine script content AppendLine body content AppendLine lt html gt File WriteAllText fileName content ToString Therefore the server side module processed the request from the client side module which asked to view the speedGraph relative to the data selected in the previous section 6 5 1 and transmitted the file Html to the client side module 6 5 3 2 Client side The client side module of the software was designed with the aim to send to the server side module requests made by t
194. y the RTLS in the same free stall barn where RTLS performances were evaluated was the ultimate goal of this thesis The purpose of this test was to assess the performances of the implemented features and to verify the effectiveness of the results obtained from the use of the tool in order to find useful informations related to the occurrence of the physiological state of estrus Chapter 3 WORK ORGANIZATION Part of this thesis work shows a review of studies carried out in the field of precision livestock farming In particular ICT applications for indoor animal identification and localisation are listed in order to illustrate the state of the art in the introduction of new approaches based on the use of high tech automated monitoring systems in the study of animal behaviour Part III contains the materials and methods of the research In detail chapter 5 describes the properties of an Ultra Wideband Real Time Loca tion System In the last section 5 3 attention was focused in the description of the system adopted in this research Furthermore this part gives a detailed description of the case study chapter 6 carried out within of a dairy house located in the province of Ragusa Sicily Italy In particular after a de scription of the area of the barn under study section 6 1 the analysis and installation of the Ubisense UWB RTLS section 6 2 are outlined in detail Subsequently the multi camera video recording system section 6 3
195. y the end user Such requests in particular concern the viewing of the cowSpeedGraph in relation to the parameters set by the user In order to calculate the instantaneous speed relative to the data selected in section 6 5 1 the following data structures are implemented through the classes LocationEvent ScalarVelocityEvent and ScalarVelocityNDEvent In particular the LocationEvent class was implemented to contain the date tag ID and the values for the three position coordinates x y and z for each record extracted from the database The following source code is related to the class just described 1 2 o 0 A O 10 public class LocationEvent region properties public DateTime Date public string IdTag public double X public double Y public double Z endregion 6 5 SOFTWARE TOOL FOR THE VISUAL ANALYSIS 107 11 region constructors 12 internal LocationEvent 13 DateTime date 14 string idTag 15 double x 16 double y 17 double z 18 1 19 this Date date 20 this IdTag idTag 21 this X x 22 this Y y 23 this Z 2 24 25 endregion 26 The ScalarVelocityEvent class however was implemented to contain the date tag ID and distinct instantaneous velocities for each coordinate 2 y and z The following source code is related to the class just described 1 public class ScalarVelocity Event 2 3 region properties 4 pub
196. ystem which was introduced by the Spanish cattle farmers association FEVEX utilized RFID tags injected into the cows hooves to allow a quick and accurate animal data retrieval Furthermore the Canadian Cattle Identification Agency replaced barcode tags with RFID tags to iden tify a bovine s herd of origin and this was used for the traceability of packing plant meat Finally RFID tags were used by Feng et al 2013 to perform tracking and storing of dairy cattle breeding data in order to implement a real time traceability management system along cattle production flow 4 2 Animal localisation With regard to cow localisation some research studies investigated the possi bility of using different indoor positioning systems for cow tracking Huhtala et al 2007 demonstrated that a system which uses Wireless Local Area Network technology and Time Difference of Arrival TDoA algorithms pro vided a localisation error of about 1 m with a position accuracy of 7096 4 2 ANIMAL LOCALISATION 15 in undisturbed conditions no moving cows clear sights between antennae and tags etc T gersen et al 2010 used the Bluetooth wireless technol ogy to monitor dairy cows in a barn and demonstrated that is possible to locate a cow with a precision of approximately 0 6 m in stationary conditions Ipema et al 2013 obtained an average accuracy of 0 30 m with a standard deviation of 0 25 m in a static accuracy test carried out in a cattle barn by

Visualizza/apri

Contents

Download Pdf Manuals

Related Search

Related Contents