An Embedded Module for Robotized Inspection of Power
Contents
1. Mere Manual Jon 200 impt student mana pdf YE Chow and 1 Yan Automatic diagnosis system of electrical ung lnc teeogepin se Proceeding of the 2009 Intnl Conor on Sf Recon N Ow Selection tebe eon gry eel histograms pps cect TS Janay 1995 at pp etree 889 rate Working Group RFC 4866 July 2006 avaliable 3t pip sett ore LVESSS Soman Media Live Newer Inc View CA 201 up ieSSS com need LIS Li ay code ry 2012p ib The GTK Team The GEKS project 2012 tka an ua2. Trop which representa the cable temperature under normal The referee temperatune ia compuned based the foreground without the hotspots TW a Where 7 denotes the complement of The reference temperature mask has a total number of pixels given by Me LDA Then the reference temperature can be computed as Tes pE Ehei ITT Hotspots are classified according to table based on the percent raise of its average temperature With respect to the reference temperature Trey 9 Toes a gt as Tamen a 9 lt 907 MENA The determination of the threshold value for image seg mentation is the most demanding slep in terms of computing resources as the optimization is performed by exhaustive search The variances should be computed by supposing that the threshold value is each point in histogram of 16 bit data That means to compute the variances of the foreground and background 65536 times for each frame In order to reduce the computing time an auxiliary histogram is built by using only the most significant byte of each data That way a 256 time smaller histogram is obtained which is used to compute an approximation of the optimal threshold value This approximate value is used as initial point for a local search for the optimal point in the full histogram 2012 IV IMAGE RECEPTION AND TRANSMISSION The st
3. RTSP protocol The Session Description Protocol SDP 13 is used to describe the desired session parameters such as such as session name media name address and connection information The SDP protocol is used in the setup phase this phase the server describes the available streams the client which then configures the parameters for the desired stream Figure 4 shows the SDP description for the RAW 320 x 240 stream used in this work The SDP session description begins at the line which stats with v 0 SETUP PLAY Notice the line starting with a rtpmap which starts the description of the desired stream The stream is deseribed by tpmap 103 raw 90000 a framesize 103 and 103 sampling mono eight 240 depth 16 ig SDP description far the RAW 320 24 stream SOFTWARE IMPLEMENTATION The software executing in the inspection module is divided in three modules running as daemons as shown in figure 5 receives images from the thermographic detect captures images from the visual camera acquires the images from the receiver and detects host spots and faults generating a synthetic images with the hot spots and fault data transmitter Transmits the stream of synthetic im ages generaled by the detector There is not a specific daemon for the capture of visual camera images because the camera supports the UVC format Which has an available driver in the Linux However
4. the abstract classes of the liveMedia library were derived in Iwo new classes ShmSink to receive a stream through shared memory and ByteSt reamShmSource to send the stream through shared memory The ShmSink class is used by the receiver damon to receive streams from the thermographic camera through shared memory while the ShmSource class is used to to send the stream to the trans j shared memory Note that those new classes are extensions of the classes from the liveMedia library which as not changed in any way the compatibility and correctness of the stream handling The receiver daemon receives stream from the thermo graphie camera an sends the dala to the detector daemon An RTSP session with the camera is initiated then the appro priate stream RAW data without compression is selected The received RTP frames are sent to the detector daemon The frames are sent with the headers defined by RFC4175 and there is no direct correspondence between each RTP frame and each image frame The detector daemon should concatenate the frames to assembly an image frame to be processed or daemon processes the image and generates stream with the results which is transmitted by the remote monitoring station by the A requirement for the development of the inspection module Was that the monitoring station should be required to execute any proprietary sofware in or
5. thermographic camera is a FLIR A320 which send images through an Ethernet connection using RTSP Real Time Streaming Protocol 6 Table 1 shows the stream formats supported All formats are 16 bits per pixel Ee Note that the MPEGS 640 x 480 format is an interpolated resolution as the sensor resolution is 320 240 The FCAM is Flir proprietary format for raw data Streams in RAW format be either raw data from the camera sensor or a radiometric image calibrated in Kelvin The measurement range is from OPC to 350 C with an accuracy of 42 C The temperature resolution can be set to be either 0 1 Kelvin or 001 Kelvin Tn this work a 320 240 RAW radiometric stream with resolution of 0 1 Kelvin was used The frame rate is 9 The processing module is a BeagleBoard xM which uses an ARM processor due to its good ratio between processing power and power consumption which also motivates its use in cell phones PDAs and tablets Another important point is that the ARM processor is well supported by the Linux operating system which can be easly customized for a low Footprint in memory and system resources The OpenEmbedded cross compilation framework 7 is used to generate a customized version of the Linux ngstr m distribution 8 The visible spectrum camera is Logitech C2010 which uses an USB After the processing of the theemographic and visual im ages the embedded inspection module generates another RTSP st
6. 2nd International Conference on Applied Robotics for the Power Industry ETH Zurich Switzerland September 11 13 2012 2 An Embedded Module for Robotized Inspection of Power Lines by Using Thermographic and Visual Images Walter Fetter Lages Department of Electrical Engineering Federal University of Rio Grande do Sul orto Alegre RS 90035 190 BRAZIL E mail feter ece ufrs br Abstract This paper deals with an embedded module for automation of thermographic inspection The module captures ideo streams from an infrared camera and from a visible image Camera and performs an image processing to detect faults The Protocols used to capture the Image streams and the image Processing are discussed The results with the faults highlighted fre sent through synthetic image stream to supervision 1 INTRODUCTION Power lines can wear out because of several factors which can result in high power loss or even complete power faults caused by broken lines Emergency repair procedures are high costly motivating the development of systems capable of detecting damaged lines in a predictive manner Power line inspection robots have been proposed to overcame the problems associated to human inspection such as the tedious task and the resulting fatigue of operators This paper proposes an embedded module for robotized inspection of power lines by using thermographie and visual images That module receives images from a the
7. der to enable any computer to be used as monitoring station Therefore the stream with results smitter daemon is transmitted in MPEG format which can be visualized by using any conventional player such VLC or Mplayer The conversion of the video stream to the MPEGS format is done by using the Likavcodec library 15 Even if the processed stream is transmitted to the remote monitoring station for debugging purposes it is convenient to view the images while they are still in the processing module Hence the detector has a optional graphical interface based on the GTK library 16 which can be used to see the generated stream before transmission to the monitoring station Figure 6 shows such an interface The upper left corner of the Figure shows the thermographie image while the lower left images shows the detected foreground The upper right image shows the detected hotspot and the lower right image an image in the visible spectrum The text in the bottom of the window shows the classification of the hot spot VI Coxctusto Experimental results are shown in figure 7 where a the screen of the monitoring computer running the VLC player be seen A damage in the cable was mechanically forced thus reducing its section high current power source was used to simulate the high currents seen real power lines The system was able to efficiently detect faults producir real time report for the operat
8. ground pixel values mean of all pixel values Given the thermographie image a r the foreground alz y is given by if Alay 1 au otherwise Where the foreground mask r y is given by OEE trates Then the hot spots are detected based the hottest pixels of the foreground First the hottest pixels in Foreground image are searched for resulting in a seed image mng Sgr o The hotspots are obtained by growing the seed im age Qy x y in recursive dilation operations with an 8 neighborhood mask B Hence a neighbor pixel belongs to the hot spot if its tem perature is greater than 0 6 Tmas Toun Dilation op erations are recursively performed until convergence to stable hot spot regions such that a z y fleir y 07502 0 Small hot spots are regarded as a consequence of noise and are discarded Therefore the dilation operations makes the method more robust and less sensitive to noise as the hot spot detection requires local support of many pixels Furthermore the dilation avoids the segmentation of a hot spot in iwo or more due to noise Each connected region in is a hot spot with i 16 The mean temperature of each hotspot Ther is used 10 detect faults in the cable and is given by LEE lew Maw Adley with ay Bach hetpot temperature Tha i compared mean temperature the the detected hot apais
9. or where the detected faults highlighted The requirements of low power consumption and small dimensions were satisfied as well ACKNOWLEDGMENT The authors would like to thank the financial support from Conselho Nacional de Desenvolvimento Cient fico 2012 Tecnol gico CNPq Coordena o de Aperfei oamento de Pessoal de N vel Superior CAPES Funda o de Apoio Pesquisa do Estado do Rio Grande do Sul FAPERGS and Companhia Estadual de Energia El trica CEEE REFERENCES LU 8 Abou tbe future of power line robotics in Proce int of the 1 ent Conference Ap Rts jor he de Oia and W F Laget Control of a beachiation robot oc Inston of aerial power nest Proceding of the st Conference on Appied Robotics for the Power industry Most 2010 H Li and Wang Reseach on and wale pover image for inspection robot in Proceedings of the tran Cine Spied abo for Pray H E de Oliveira and W Lages Robodzed impection of power limes ws infrared vision in Proceedings ofthe Tantra ingen Aid Rabi for the Pover Not 2010 inspection robats Design ofthe power spy stem Proceedings ofthe tt Internationa Conference Ad Robotics for he Poner m i 2012 7 Opeaintedied Team Opentbeded User 2009 pes
10. ream from the thermographic camera isin the format defined by RFC4175 RTP Payload Format for Uncompressed Video 11 The uncompressed format is ideal for image processing as there are not distortions introduced by the compression and there is no time spent decompressing the image After the image processing and fault detection the inspection module transmits the results using the MPEGS format which is a compressed format in order to reduce the required bandwidth and to enable the visualization in any standard media player such as VLC or Mplayer The results from the inspection module are used just for visualization and not for processing Hence there is no need to transmit uncompressed video However in both cases the stream is transmitted as RTP Real Time Protocol 12 frame Figure 2 shows the protocol stack a TCP 1 I The Real Time Streaming Protocol RTSP 6 is used to control the transmission of the streams through the commands OPTIONS list the supported optional commands TBE Hist the streams available by the camera UP set one RTP session using a specific stream TPARAMETER get parameters ike frame rate and format PLAY start to send the streams PAUSE pause the sending of streams close the RTP session Hence the user contol the stream transmission much like a DVD player Figure 3 shows the state machine of the
11. ream with the results of the image processing as shown in figure 1 where the detected faults and their locations are evidenced That stream is received in a monitoring station The LiveSSS library is used to receive and transmit RTSP streams s THERMOGRAPHIC IMAGE PROCESSING Each frame from the thermographic camera is processed to detect faults on the transmission line The image processing is based on the Thermography Anomaly Detection Algorithm 9 to detect hot spots and classify the fault The algorithm is based on statistical and morpholo ical methods The first step is to segment the image in lower and high temperature areas named the background and the foreground of the image respectively Only the foreground is further processed in search for hotspots since it represents the power line The background is discarded as it represents the surrounding environment and not the power line The image segmentation is done by thresholding The threshold value 17 is determined dynamically for each frame by using the Otsu method 10 which maximizes the Variance between the 2012 pixels belonging to the foreground and those belonging to the background pe me w argmaxoty oi probability of a pixel belonging to the background iy probability of a pixel belonging to the foreground Hi mean of background pixel values liy mean of fore
12. rmographie camera and from a visible image camera process them to detect faults and send the images and diagnostics to a remote monitoring station Robotic systems for power line inspections have been proposed with research focusing on the locomotion system 1 2 as well as on methods for fault detection 3 4 As the proposed module will be embedded in a robot dimensional Weight and power consumption requirements become very important It is a paradox but even with the robot operating in active lines the power available to the robot is limited since it should operate from batteries or power obtained by induction from the power line itself However the methods to obtain power from the line itself can provide only about hundred Watts 5 This paper is organized as follows in section I the hardware for thermographic and visual image capture and processing is described Section II presents the processing of the thermographie images to detect hot spots and abnormal conditions in the power line while section IV shows the details of the capture of the video streams from the thermographi 2012 Vinicius Scheeren Department of Electrical Engineering Federal University of Rio Grande do Sul orto Alegre RS 90035 190 BRAZIL E mail vscheeren ece ufrgs br camera The software implementation of the system is de scribed in section V and section VI presents some results and concluding remarks TL HARDWARE The
13. the camera image is represented in the YUV model The image is converted to the RGB model by 169 18 R 1164Y 16 1 596 7 128 16 G 1161 Y 16 OS13 V 128 0 128 an B LIGY 16 201800 128 18 2012 Receiver m m Detector l Transmitter a UVC Driver Bes Das ow The modular system architecture allows for an easy replace ment of the receiver or transmitter modules should another stream format be more convenient The communica tion and data transfer between the modules is done by shared memory thus avoiding to copy images The read and write operations are synchronized by semaphores The liveMedia library from the liveSSS project 14 is used to decode the RTP Real Time Protocol 12 received stream and to encode the RTP stream to be transmitted The architec ture of this library defines the abstract Framed i 1eSou and Mediasink classes representing a generic source or sink of streams However in the implementation of the liveSS library those classes are derived to implement only the ByteStreanFileSource and FileSink classes Those classes implement stream read and write operations from magnetic media files Due to the low performance limited by the file system that i not a good solution for sharing streams between tightly coupled tasks as the implemented by the inspection module daemons In order use shared memory based communication
Download Pdf Manuals
Related Search