Home

View/Open - Oregon State University

1. 684714 945 0 0 0 999981 0 00390492 3953516 019 9953476 108 JJF tol 2933399 315 3953369 184 3953360 725 3953374 677 0 00452797 0 00388298 0 999982 24 270 24 269 Figure A3 Adjustment results will be shown on a screen similar to this one Rotation angles are in radians and Degrees Minutes Seconds Copy and paste this output to a blank text file if you want to save it The results printed are fairly straightforward and can be copied to your favorite text editor or spreadsheet to be saved RMS is a report of the Root Mean Square error which is a measure of the quality of the match between the input data sets Usually no more than 10 iterations are required to find a match If there are more there may be issues with the adjustment S is the scale parameter Tx Ty Fz is the translation vector Rx Ry Rz are the rotation angles in radians or DMS The rotation matrix is computed from Rx Ry Rz see equations in Silvia 2011 The new points are the transformed points from the measured point file after the final iteration Page 84 APPENDIX B NISLT MATLAB SCRIPT NOTES Page 85 NISLT Matlab User Manual 5 19 2011 By Evon Silvia Graduate Research Assistant School of Civil and Construction Engineering evon silvia mail com Introduction The adjustHan ctrlPath measPath MATLAB script performs an optimized version of the similarity NISLT adjustment as developed and presented by Jen Yu H
2. December 2006 1310 1314 Page 78 APPENDICES Page 79 APPENDIX A LEAST SQUARES ANALYSIS PROGRAM NOTES Page 80 Least Squares User Manual 5 19 2011 By Evon Silvia Graduate Research Assistant School of Civil and Construction Engineering evon silvia qmail com Introduction The Least Squares program is a simple Qt interface that receives a set of matching point pairs and performs a traditional un weighted nonlinear least squares adjustment and reports the results The adjustnonlinear C class was written to be easily integrated into other programs if desired Application of the resultant transformation to additional points must be done manually by the user Data preparation Input files must be comma delimited ASCII files with X Y Z coordinates followed by a newline character on every line Comment lines must begin with a character The control and measured data must have the same number of points with corresponding points following the identical order in both input files The file must also end with a newline character a blank line General procedure For details regarding the nonlinear least squares solution see Evon Silvia s thesis Overcoming the Level Bubble Dynamic Terrestrial Laser Scanning Reference Frame Transformations 2011 Oregon State University or the textbook Adjustment Computations Spatial Data Analysis by Charles Ghilani 5 Edition 2010 Upon opening the prog
3. 0 cc ccc cccccccccccccccccccccceccccccceccenecceaeuuenceneceeaegueaesneaeecesueatsneaesneass 72 TABLE OF CONTENTS Continued Page yA 6 0121 9 0 C12 Sacra ee ee ee ee ne ee Tee 78 Appendix A Least Squares Analysis Program Notes ccccceceeeeees 19 Appendix B NISLT MatLab Script Notes c cc cccceccseceeeeeeeeeeeseeeaes 84 Appendix C Scanner Interface Program Notes ccccccceeeceeeeeeeeeeeaes 87 Appendix D Results of a Survey of TLS Professionals c c0ceees 95 LIST OF FIGURES Figure Figure 1 Scanners rotate horizontally and vertically y recording thousands of points per SECONG cccceccecceeceeeceeeeeeeeeeeeeeeeeeeeeees Figure 2 Laser scanners are line of sight instruments resulting in occlusions This soil pit was scanned from one left and three right scan positions filling occlusions red arrows using MUINIDIE SCal POSITIONS sensor sracceaeatereentatata arian te atannaeneaees Figure 3 A right handed coordinate system with positive rotations SAE a LER TEE TETEE E AE E TEE E T E E E EEEE Figure 4 Test 2 plot of TP residuals computed from tiepoint registration against scan duration on a logarithmic scale Note that all four plots appear to converge to zero at approximately 83 seconds Figure 5 Radial plots of roll top and pitch bottom measurements radial axis in degrees from level at approximate horizontal
4. Calc Duration Note that actual scan time varies from this value Scan options Image Capture The scanner has a calibrated camera mounted on top of it for collecting images aligned with the scan just taken Enable this option to have the camera collect a full set of images The number of images taken depends on the CAM_OVERLAP property which for now must be set using Telnet or RISCAN PRO There is an option to capture image sets for HDR processing This feature is still under development Use at your own risk Real Time Mode This feature will enable additional features that are needed for real time analysis Specifically the scanner will record highly accurate inclination values prior to scanning a low resolution monitor file of the scan will be downloaded to the same directory as the executable and a text file vz_info txt will be output in the same directory as the executable by default that can be parsed by a real time analysis program see below Go Click Begin Scan to send all of the settings to the scanner Verify that the settings were sent correctly by monitoring the text output in the Info log Once the scanner has begun scanning you are free to change the settings in preparation for the next scan Each new scan in a project opens a new scanposition The scan can be aborted at any time by clicking Abort next to the progress bar Scan retrieval The scanner produces several files with each scan all of
5. dS c5 edS eSd TdS c5SD wi Szeulpsoo X UGIeYS JE 0L T53 edS TSI dS TSD TdS T53 w Szeulps007 X JE AS Z 100 4 o wis Uo1qeAa J 10044 Page 66 dS uonisod ueds Aq pajeiedas 100 ML y JO S9 SUOI D9S SSOJD UJNOS UWION GL aunbi4 UOHEYS F 0OL PFSD EdS PSD g dS t53 TdS UGE FIOL ESD EdS ESD e dS ESD TdS ESD wu SJEUIPICOD A ule Aa y 1900 4 uso Page 67 4 2 REAL TIME CHANGE DETECTION Because of its flexibility soeed and efficiency terrestrial laser scanning has shown promise as a valuable tool to efficiently detect and measure change in real time Figure 16 shows an example of post processed change detection applied to a project in the HWRL Similarly Figure 17 shows a gusset plate scanned during load testing with change detection relative to a planar surface Color coding similar to what is seen in either figure can be implemented in real time for rapid user feedback By combining change detection algorithms with the surface matching azimuth adjustment technique developed by Olsen et al 2009 2011 future researchers will be able to detect change using a mobile computer in the field rather than the traditional workflow of returning to the office to process data for change analysis That augmented reality data could be used to immediately determine areas that require scanning at higher resolutions reducing unnecessary data colle
6. do makes the ILRIS the best overall choice but still 9 22 10 10 33AM view its lack of a dual axis compensator as a negative Coming from a survey background control and the use a compensators 9 22 10 7 54AM are the critical factors for scanning projects 9 22 10 5 59PM Page 103 The Topcon scanner and software is by far one of the most accurate and 9 29 10 4 43AM easy to use time of flight scanners on the market Control is the key to any good scan survey It is amazing how lack of 9 22 10 4 39AM knowledge and understanding can ruin good intentions The HDS3000 requires 4 control points each in a sector surrounding the scanner for every setup otherwise the overlapping data between scans shows a separation that becomes very difficult to deal with The sun is really hot in South Africa so we often experience changes in the tilt of the scanner due to the sun being on one side and regular target checks often reveal horrendous changes in the elevation of targets at 50 100m We recently used 2 C10 s for a highway scanning project We are still getting into mobile scanning and never experienced any separation issues even using only 2 control targets at times We had no sun issues either Very impressed with the compensator for this type of work but understand the issues behind the speed of the laser vs the rate of compensation and that the compensator can hinder the laser acquisition rate 9 21 10 7 55PM In case you did not mean
7. response not required Count 40 I m not sure 0 Every time 22 Most of the time 11 Occasionally 3 Rarely 2 Never 2 Note 1 respondent said he used his compensator every time but he entered in the previous question that he doesn t have one altered the Every Time category from 23 to 22 to reflect that discrepancy Question 4 Why do you use or not use a level compensator response not required Count 37 Stopped using it because it slows down the scan speed and takes to long to establish control for the scan setups Didn t notice any difference in 9 28 10 12 53AM accuracy We always scan level with no need for inverted scanners etc Why wouldn t you use a compensator Relying on targets alone seems a step backwards The standard deviations in height for targets after registration gives you a good QA check on level consistency in your scan cloud It does not have any negative effect but supports registration a lot 9 24 10 7 32PM We use it during scanning but don t always use it for registration depends 9 24 10 4 07AM on the residuals We use it because it is available 9 24 10 1 47AM Orientation is better defined through a control network 9 24 10 1 25AM 9 26 10 5 32PM Page 98 Because want levelled data pretty much 100 of the time as the work is l survey grade wouldn t turn off a DAC on a TPS 9 23 10 6 29PM The introduction of the dual axis compensator was a tremendous advancement in the emerging hardw
8. 0 00527 0 00102 0 99999 1 00000 0 00008 0 00012 1 00000 1 00000 1 00000 1 00000 1 00000 1 00000 1 00000 1 00000 R LS RMSE 0 1292 R LS RMSE 0 0221 0 40688 0 91348 0 00123 1 00000 0 40788 0 91304 0 00033 1 00000 0 91346 0 40689 0 00529 1 00000 0 91304 0 40788 0 00054 1 00000 0 00533 0 00103 0 99999 3 1 00000 0 00063 0 00008 1 00000 1 00000 1 00000 1 00000 1 00000 1 00000 1 00000 1 00000 Table 2 Comparison of roll pitch and yaw values derived from rotation matrices produced by different adjustments with and without outliers sori soon Point Santon TTR mod ms R amoa mis Roll 2 6481 0 0583 0 0591 0 0764 0 0069 0 0046 Pitch 0 8787 0 3022 0 3053 0 0213 0 0045 0 0359 Yaw 65 9298 65 9657 65 9904 65 9189 65 9261 65 9283 Page 23 2 3 TRANSFORMATION AND ADJUSTMENT CONCLUSIONS Han 2010b demonstrated that the NISLT technique could be reliable for high precision synthetic data producing results similar to traditional least squares methods in a fraction of the time Nevertheless caution should be exercised while using the results of the NISLT technique because rotation matrix properties are not conserved It is interesting to note that the RMSE reported using R is lower than the least squares adjustment R LS which should by definition produce the result with the theoretical minimum RMSE The rotation matrix approximation R do
9. AHMCT Final Report California AHMCT Program University of California at Davis 2007 RIEGL LMS GmbH RISCAN PRO v1 5 7 May 18 2011 Page 6 Salvi Joaquim Carles Matabosch David Fofi and Josep Forest A review of recent range image registration methods Image and Vision Computing 25 no 5 May 2007 578 596 Skaloud Jan Philipp Schaer Yannick Stebler and Phillip Tome Real time registration of airborne laser data with sub decimeter accuracy SPRS Journal of Photogrammetry and Remote Sensing 65 no 2 March 2010 208 217 son Seokbae Hyunpung Park and Kwan H Lee Automated laser scanning system for reverse engineering and inspection International Journal of Machine Tools and Manufacture 42 no 8 June 2002 889 897 Vosselman George and Hans Gerd Maas Airborne and Terrestrial Laser Scanning North American Caithness Scotland Whittles Publishing 2010 Vuylsteker Pascal Computer Graphics 3D Transformation 14 19 Properties of the Rotation Matrix April 20 2004 http escience anu edu au lecture cg Transformation rotationMatrixProp en ht ml accessed May 18 2011 Wehr Aloysius and Uwe Lohr Airborne laser scanning an introduction and overview ISPRS Journal of Photogrammetry amp Remote Sensing 54 1999 68 82 Page Wynne Randolph H Lidar Remote Sensing of Forest Resources at the Scale of Management Photogrammetric Engineering amp Remote Sensing ASPRS
10. Every time you press a button the scanner s setup has been disturbed a small amount which can be a problem in high precision work e RISCAN PRO Every laptop and computer in the lab has RISCAN PRO installed which can run the scanner either wirelessly or through an Ethernet connection This program can open display edit and process the raw data produced by the scanner It can also conveniently record new scans based on the previous scans e Telnet The scanner can be connected to using telnet a command line communication protocol This sends raw commands that the scanner interprets A list of these commands and their use is in the scanner s manual RISCAN PRO and DriveVZ400 both use these commands in a manner invisible to the user Connecting to the scanner Open the executable either by double clicking it or running it from the command line see the next section for instructions using the script codes The interface should appear Page 89 VZ400 Driver ee t Project 20110101 _FieldScanning d T Debug Storage Media USB Internal auto Measurement Program Atmospheric Conditions Temp 12 70 H Pr 1000mbar RH 60 gt Scan Pattern settings Use preset O Manually enter Vertical From 30 000 H To 130 000 H Res 0 09875 Horizontal From 0 000 To 360 000 Res 0 40000 Capture Images HDR Images Real Time Mode Seon So Project list download successful Figure C1
11. LIST OF FIGURES Continued Figure Page Figure 17 A gusset plate scanned before and after load testing colored according to deviations from a plane Red indicates large GISPIACEIMIEINS serrie iaa 68 LIST OF TABLES Table Page Table 1 Comparison of 3x3 rotation matrices from different adjustment methods for a set of GPS points with and without outliers Italicized values are the sum of the squares of the corresponding FOW OF COUN oenar siaiacatgnaiutuiasanaeetstaiabatetatebetaimneataiatataisies 22 Table 2 Comparison of roll pitch and yaw values derived from rotation matrices produced by different adjustments with and without SEE Seacoast ener etn acer sect actrees ete tte ara E EAE AAE ae 22 Table 3 Summary of inclination sensor measurement techniques of four example scanners Modified from JADA Productions 2010 FARO Technologies Inc 2009 and Leica Geosystems AG 2011 35 Table 4 Test 2 results for 360 scans from both unlevel and level scans DM roll and pitch values measured using the discrete method described in Table 3 SW roll and pitch values extracted from the raw scan data using the scanner s ACCOMPANYING SOMWaAN lE ssaptire nE ate aceacetoashenetcacuaniadals 39 Table 5 Comparison of inclination measurement methods in degrees cw Clockwise and cew counter CloCkWISe c sscesseeeeeeeeees 46 Table 6 Comparison of inclination data from inclination sensors com
12. developed by Han 2010b from eigentheory has shown promise as a reliable and efficient alternative to the sometimes computationally intensive nonlinear least squares solution Rather than simultaneously distributing the overall error evenly between the seven transformation parameters as in least squares the NISLT technique computes an average scale rotation and translation each independently This section summarizes the technique itself as presented in Han 2010b Section 2 2 4 presents optimizations of the technique while section 2 2 5 contrasts NISLT against least squares Given the matrix form of the similarity transformation equation in 2 7 and point coordinates in the vertical vector form of 2 1 scale can be estimated by comparing the distances between all control point combinations to the distances between all original point combinations and taking an average mean fel i j 2 10 lax where dx is the distance between the original points i and j and dx is the distance between control points and j The rotation matrix R can be estimated using an equation developed through eigentheory R AX AX AX AX 2 11 Page 19 where AX is formed by stacking the transposed and scaled vectors between the original points and AX is formed by stacking the transposed vectors between transformed points as shown dxi ax T IT AX ee AX e 2 12 T T a ixj any ixj Finally the tra
13. 0 007 0 719 0 002 1 691 0 005 1 691 0 005 185 0 04f 0717 0 004 0 722 0 001 __ 1 687 0 009 1 691 0 005 8 oof 0714 0 007 0 721 _ 0 000f _ 1 636 0 010 __ 1 688 0 003 6 war hos ora oo 0 728 0007 1646 0 050 1 709 0 013 average 0 721 0 712 standard deviation 0 008 0 024 iepoint registration 0 721 EVELED Time Elapsed Resolution sec DM res sw 740 0 02f 0 014 0 00 475 0 025 0 006 0 004 0 009 _ 0 007 0 002 __ 0 047 0 007 __ 0 036 0 004 piss 004 8B 0 _nodataj 0014 _ 0 009 _ no dataj 0 050 0 010 0 0 09875 _ 0 003 v0 1 h0 5 f 0 004 0 53 average 0 039 0 018 standard deviation 0 011 0 180 iepoint registration Ss 0 040 Page 40 Legend TP Roll Unlevel E TP Pitch Unlevel TP Roll Level TP Pitch Level a N U a an Q A a4 Scan Duration seconds Figure 4 Test 2 plot of TP residuals computed from tiepoint registration against scan duration on a logarithmic scale Note that all four plots appear to converge to zero at approximately 83 seconds It had been observed in previous versions of this test that the simple act of nudging the power cable resulted in instantaneous variations of roll and pitch by as much as 0 10 By operating the instrument remotely and eliminating all physical contact with the scanner upon setup completion the scans were noticeably m
14. 21 2009 http www faro com accessed November 30 2010 Ghilani Charles D Adjustment Computations Spatial Data Analysis 5 Hoboken NJ John Wiley amp Sons Inc 2010 Han Jen Yu A Noniterative Approach for the Quick Alignment of Multistation Unregistered LIDAR Point Clouds EEE Geoscience and Remote Sensing Letters 7 no 4 October 2010 727 730 Han Jen Yu Noniterative Approach for Solving the Indirect Problems of Linear Reference Frame Transformations ASCE Journal of Surveying Engineering 136 no 4 November 2010 150 156 Han Jen Yu Jenny Guo and Jun Yun Chou A Direct Determination of the Orientation Parameters in the Collinearity Equations EEE Geoscience and Remote Sensing Letters 8 no 2 March 2011 313 316 Page 74 JADA Productions Laser Scanner Hardware Survey Point of Beginning Edited by Christine ts Grahl 2010 http laser jadaproductions net nardware default php accessed November 15 2010 Jekeli Christopher Inertial Navigation Systems with Geodetic Applications Berlin de Gruyter 2001 Leica Geosystems AG Leica ScanStation C10 Technical Datasheet English US 2011 http nds leica geosystems com en Leica ScanStation C10 79411 htm accessed February 24 2011 Lichti Derek D Error modelling calibration and analysis of an AM CW terrestrial laser scanner system SPRS Journal of Photogrammetry amp Remote Sensing 61 2007 307 324 Lichti Dere
15. 3 y where R is the combined 3x3 transformation matrix of the three axial rotation matrices and Ppr is the point vector in the rotated reference frame Note that matrix multiplication is not commutative so R RxR Rz would result in a different rotation matrix Multiplying the matrices together in order from left to right such that fm is the element of R corresponding to row n and column nm produces M1 2 3 R 1T21 122 123 2 4 r31 132 133 11 cos 0 cos 0 12 cos 6 sin 0 sin 0 sin 6y cos 6 13 sin 0 sin 6 cos 0 sin 6y cos 6 11 cos 0 sin 6 T22 cos 6 cos 0 sin 0 sin 0 sin 6 Sy WwW sin 6 cos 6 cos 0 sin 6y sin 0 T3241 sin O T32 sin Ox cos Oy T33 COS 0 cos By Page 11 Roll pitch and yaw can be extracted from the rotation matrix in 2 4 by arctan r33 Ox 0 arctan 4 2 5 y 0 ritr arctan T11 A rotation matrix is a special orthogonal matrix and as a result must demonstrate certain properties Vuylsteker 2004 1 The inverse of Ris equal to the transpose R R 2 The determinant of R 1 3 The sum of the squares for any row or column in R equals 1 4 The dot product of any pair of rows or any pair of columns in R is 0 5 The rows of R represent the unit vector components in the original reference frame along the axes of the rotated reference frame 6 The columns of R represent the unit
16. ACKNOWLEDGEMENTS Partial funding for this research was provided through OTREC under project 2011 398 The authors would like to thank Leica Geosystems and David Evans amp Associates for their generous provision of training equipment and software to Oregon State University Maptek Pty donated licenses and support for Maptek I Site Studio Riegl Laser Measurement Systems provided their RiVLib API and technical support for the Riegl VZ 400 Shawn Butcher OSU CCE doctoral student and Tim Maddux HWRL assisted with data collection and background information Finally Gene Roe of lidarnews com and administrators of laserscanning uk helped distribute the inclination sensor survey We appreciate those who participated in the survey and shared their unique insights Page 5 4 FIELD APPLICATIONS OF TLS REFERENCE FRAME TRANSFORMATIONS Surveyors and geomatics professionals provide a valuable service as measurement and spatial data modeling experts Several collaborative efforts with other members of the Oregon State University research community were undertaken for training database enrichment and field testing Projects related to research conducted at the O H Hinsdale Wave Research Laboratory HWRL on the Oregon State University campus will be highlighted in this section 4 1 THE HWRL TSUNAMI WAVE BASIN COORDINATE SYSTEM The need for a new local coordinate system in the HWRL Tsunami Wave Basin TWB was first introduced in Section 3 4
17. Dont think knew about that 9 28 10 12 53AM Not when using the Leica 6100 as we are only relying on the bubble When scanning on our extendable tripod up to 3 metres in height we may not use the compensator if confident the setup is stable in this case we 9 26 10 5 32PM rely on targets as it is often difficult to physically level the scan head to within compensator tolerance Page 101 If the residuals are better with the compensator turned off then we visually inspect the scan for double clouding and if there is none then we leave the 9 24 10 4 07AM compensator off some times the compensator does not work well when stiching the scans 9 24 10 1 47AM together In that case we will disable it use Faro s inclinometer ON setting to register our scans 9 22 10 7 24PM Only if control is not available in quantity or quality 9 22 10 4 39AM Not during high accuracy scanning in which a large number of control targets are used and the z axis of the scanner is not able to be plummed Depands on the solution of registration and site conditions 9 21 10 5 52PM The Leica compensator is either on or off If off then we switch to a more complete cloud to cloud or multitarget system If it s on then we treat the clouds as levelled The compensator used is the same as in the Leica total stations 9 21 10 10 26PM 9 21 10 5 09PM yes when using a scanner with a DAC and never if it doesn t even when the software wants to think that slig
18. a minimum recommended scan time of 100 seconds 2 minutes to be conservative balances the need for efficient field work with reliable inclination sensor data Note that the same results were achieved regardless of whether or not the instrument was precisely leveled These tests verified both the accuracy and precision of the laser scanner s inclination sensor provided that the scanner rotates slowly enough for measurements taken during scanning It is recommended that those who rely on inclination data for control should perform a similar stability threshold evaluation for their scanner to ensure that they achieve reliable results Alternatively inclination data derived from a method similar to the continuous measurement method shown in Table 3 where inclination measurements are averaged from 360 rotations clockwise and counter clockwise has also shown itself to be reliable The case study presented herein has demonstrated that inclination sensors can be a reliable and inexpensive means to verify control coordinates and scan data The additional redundancy provided by the inclination sensors and using more than the minimum control prevented propagation of the leveling errors encountered in this case study It must be emphasized that while laser scanning may be considered different from traditional surveying it must still Page 56 follow the same principles of redundancy and systematic procedures to avoid major blunders 3 6
19. an ordinary inclinometer but a dual axis compensator this is not included in the ILRIS Nevertheless the highest accuracy you can get Is by aligning all the scans via ICP and later georeference via GPS coordinates for each origin of the scan s coordinate 9 21 10 5 31PM system As the ILRIS has a very precise inherent implementation of a measurement coordinate system a compensator is not needed Dual axis compensators are needed for one mirror and rotating base systems Your original question is flawed a compensator and inclinometer are not the same thing you need to be careful how you interpreted the 2 results depending on the scanner in use lecture I am not a teacher the use of scanning and am appalled all the time to find surveyors do not know what 9 21 10 4 26PM a compensator is or what it can do for them virtually all registrations get to see from non compensated lots scanners normally have issues and are poorly registered feel free to contact with question Joseph Chumbley Lockheed Martin 9 11 10 4 28AM XXX XXX XXXX In large works we always assure the targets with traversing w total station 9 3 10 11 44PM This eliminates any risk of non closure SS2 with compensator is much slower so we sometime start with comp 9 3 10 10 48PM and turn it off later Previously have not used a compensator when using the SS2 in the roof space of a old building The roof had very limited clearance and was unsafe in places so was
20. and compute average Z value 98 471841m Adjust control by 98 471841m in Z Align scans to updated control points direction Use least squares methods to align scans to control points Verify proper alignment of scans between scan positions Align scans to updated control points v Scans did not align in X direction Isolate the wavemaker face in combined scans and fit a plane to it Force the plane to be vertical since scans were level Surficial variations of the wavemaker face introduced error to the plane fitting algorithm of 0 0135m Adjust the control by 0 0135m in the X direction Compute the normal vector of Align scans to updated control points the plane 1 000000 0 000689 0 000000 to find the Z rotation y such that the wavemaker face is Verify agreement parallel to the Y axis between total station 179 960523 and scan floor data aeee eaae Total station data consistently axis by 179 960523 using the below scan data plane s origin 144 105m 86 381m 1 066m as the center Conduct numerical least squares of rotation analysis in Z direction to minimize error between total station and scan data Adjust Align scans to updated control points control 0 0054m in Z direction to correct for range biases M Figure 11 Flowchart detailing the steps to
21. be conceptually similar their uses as described above in the Scan Registration section and implementation in TLS are far more complex In the case of a total station the instrument remains static during each discrete point measurement However many TLS systems simultaneously rotate and record thousands of points in the same amount of time as a single point from a total station This manuscript will show that this distinction between static and dynamic inclination data collection is significant particularly since some high speed laser scanners allow setups as much as 15 degrees off center to be corrected to level whereas total stations normally force a near level setup prior to data acquisition Additionally because of the wide variety of applications for TLS scans may be collected by individuals without a surveying background who may not understand the implications of data being out of level This manuscript presents an analysis of the impact of scanner rotation speed on inclination sensor results from several test procedures Additionally a case study undertaken at the Hinsdale Wave Research Laboratory at Oregon State Page 34 University that demonstrates the value of quality inclination sensor data is also presented 3 3 INCLINATION SENSOR ACCURACY STUDY As described above in the introductory sections of this paper and will be demonstrated in the case study presented in the next section TLS inclination sensors have significa
22. be found If redundant data with errors is collected multiple solutions exist A solution that minimizes the sum of the square of the errors can be computed ATAX ATL 2 8 where A is a stacked matrix of the coefficients of the linear equations X is a vertical vector of the unknowns and L is a vertical vector containing the right hand side of the linear equations Solving for X will give the solution X AlA 1A L 2 8b Weighting can also be introduced to each measurement using the diagonal matrix W often the inverse of the standard deviation of the measurements Page 16 X AWA tATWL 2 8c However if the equations are nonlinear they must be linearized prior to implementation in the least squares solution often using a first order Taylor series approximation Finding a solution becomes an iterative process that must begin with approximations of the final solution X and each iteration computes corrections e to be applied to the previous guess In matrix form e J wJ J WK 2 9 where OF OF OF dx oka OX ee Le F X Xo dx oh 70k ae l F x 4 x Xn e J Ox 7 dx K 2n in AX TA 2E aE LU Fy 1 Xa Xn OX OX OXn W is the weighting matrix as in 2 8c F through Fm is the set of original normal equations J is a Jacobian Matrix of the partial derivatives of the normal equations and through m are the elements from L in 2 8c With each iteration e is added to X and the pro
23. counter clockwise rotations A scan duration within a stability threshold can also minimize the inertial effects and variability in inclination sensor readings for scanners continuously measuring inclination data during rotation Finally inclination sensor readings from a full 360 rotation are recommended to reduce systematic bias Page 2 3 2 INTRODUCTION Terrestrial laser scanners TLS continue to grow in popularity as increasingly more applications are discovered Vosselman and Maas 2010 present a detailed overview of TLS and airborne laser scanning technology related errors potential applications and data processing techniques while Wehr and Lohr 1999 provide an overview of laser scanning in general As the TLS industry develops it will likely continue to become less expensive while enabling more rapid data acquisition following the trend already seen in airborne laser scanning Wynne 2006 Although scan point clouds are undeniably impressive to both the casual observer and geomatics professional TLS technology is still in its infancy and thus its precision and accuracy still require evaluation These studies are particularly important given the complexity of error sources native to laser scanning Attention to TLS error and calibration has increased with developments in the technology Lichti 2007 emphasizing impacts from angular displacement mixed pixels detector saturation blooming multipath and incidence a
24. scanner can quickly cause the instrument to become out of level As the target based control is acquired by the scanner at the beginning of the scan session the compensators adjustments ensure a reasonable amount of accuracy throughout entire scanning process 9 4 10 6 11AM With the compensator on we can traverse with the scanners and only l need 2 tragets at the backsight and foresight AE EAN Tppography infrastructure long range lt 100m yes other apps seldem 9 3 10 10 48PM a Setup methods like known backsight or traversing require a levelled instrument b Scans of longer duration Compensator compensates slight 9 3 10 2 35PM movements due to sunlight or other effects Using the scan station 2 i would not use the compenstor in awkward situations where a using a tripod is not possible or practical however occasionally it may be used for a time saving reasons where there is only a limited window Whereas because the c10 can be levelled using the tri brach it will always be used levelled as the speed of it usually negates the need to turn it off 9 3 10 3 15AM Question 5 Page 100 Do you use target based registration or some other technique to georeference and combine scans multiple selections allowed Count 104 I m not sure 0 Target based registration 46 Software based point cloud surface matching e g Iterative Closest Point DA ICP Manually pick common points from point cloud 21 Other please specif 13
25. that they rely on their level data to provide additional control reducing field time required to set up additional targets thereby simplifying scan registration to an existing coordinate system georeferencing The most common complaint regarding inclination sensors was that their use significantly increased scan time for certain scanner models which explains why some users choose not to always have them enabled 3 2 4 Scope of Manuscript Due to their established presence in the context of land surveying inclination sensor applications in TLS and related error have not been heavily researched Inclination sensor error is generally assumed to behave similarly to what has been shown with digital levels theodolites and total stations e it is entirely a function of plate bubble sensitivity Lichti Gordon and Tipdecho 2005 and calibration error Mechelke Kersten and Lindstaedt 2007 when the error is small Large leveling errors are generally attributed to environmental instability modeling biases and scanner resolution in addition to the TLS errors previously listed Additionally several scanners operate at close ranges Page 33 lt 100 m reducing the impact of slightly unlevel data However for scanners that can acquire data at longer ranges gt 1 km inclination sensors become increasingly important to ensure data accuracy Wehr and Lohr 1999 While the function of inclination sensors in TLS and total stations may
26. the nonlinear least squares method increases exponentially to compute the inverse a computationally intensive procedure the NISLT technique increases linearly as the data set size increases making it very attractive for LIDAR data applications in which data sets can easily number in the millions of points A typical way to measure the effectiveness of an adjustment is to calculate the root mean square error RMSE ae lay 2 IL 7 2 RMSE Ene oe Vi z z i 2 14 Ten RTK GPS points were collected near the Oregon State University campus and adjusted to control coordinates using the NISLT and traditional least Page 21 squares methods It was observed that two points suffered significantly from multi path and were removed as outliers for a second comparison The results of these adjustments are presented in Table 1 Italicized values in Table 1 are the sum of the squares of the corresponding row or column and should always equal exactly 1 according to Property 3 listed in Section 2 1 1 When 2 5 was used to extract roll pitch and yaw from the rotation matrices the results from R were significantly different from the results of a traditional least squares solution as shown in Table 2 R calc in Table 1 was computed with 2 4 from the roll pitch and yaw values extracted from R resulting in very poor RMSE values especially with outliers present R mod was computed numerically using a spreadsheet to simulate a NISLT so
27. used Multiple options have been developed to register scans together although most implement some form of a least squares adjustment to minimize the difference between common targets at known locations target registration common points cloud to cloud registration or common objects surface registration For more detail see section 3 2 1 of the thesis and Vosselman and Maas 2010 Figure 2 Laser scanners are line of sight instruments resulting in occlusions This soil pit was scanned from one left and three right scan positions filling occlusions red arrows using multiple scan positions Laser scanners generally have three different implementations each with its own advantages The simplest mode is terrestrial laser scanning TLS which entails mounting the instrument on a static setup such as a tripod where it remains for the duration of the scan Mobile scanning extends this concept Page 5 with GPS and an IMU Inertial Measurement Unit by mounting the scanner on an automobile or boat for rapidly modeling a corridor such as a highway or river without interrupting the flow of traffic or exposing field crew to vehicle traffic Barber Mills and Smith Voysey 2008 Finally airborne laser scanning ALS among geomatics professionals or simply LIDAR in other contexts takes mobile scanning into the air on a plane or helicopter for terrain or forest modeling Wehr and Lohr 1999 Point clouds produced by these
28. will always utilise the compensators Saves setting up extra targets 9 22 10 5 35AM use it to ensure that am getting the best data that can the only time have turned it off is when there is a lot of foot trafic around the tripod while 9 22 10 4 43AM indoors It is a nice backup and makes bad control easier to find Some situations limit control Using it seems to keep a real tight registration with targets 9 22 10 2 42AM Having the ability to level a scan means a higher degree of accuracy can 9 22 10 12 19AM be maintained and control for each scan If we are scanning an area that has vibration of from large truck or trains This vibration keeps making scanner switch to the next scan world We will disable the Level compensator and up sphere targets to help tie the scans together use one because a it means can be confident in any individual scan being level and b can use it as control which allows me to only use two 9 21 10 11 54PM targets in addition to this in each scan It doesnt have a level compensator 9 21 10 10 52PM 9 22 10 4 39AM 9 21 10 11 59PM Page 99 The compensator is used in most applications with the primary exception being high accuracy scanning in which a large number of control targets 9 21 10 10 26PM are used and the z axis of the scanner is not able to be plummed Do not use it when the setup might not be steady enough and has vibrations ship deck vibrating plant when compensator might giv
29. 0 003 Discrete As shown in Table 5 clockwise measurements consistently measured roll significantly higher than the average while counter clockwise measurements were low The pitch values showed the opposite pattern but with less consistency Pairing and averaging the clockwise and counter clockwise scans into ten sets of two scans improved the standard deviations and range of both roll and pitch by at least a factor of two This suggests that inertial effects can be reduced by performing scans in both the clockwise and counter clockwise directions for inclination data recorded at high speeds Recall that the listed Page 4 one sigma value of the instrument used in this study was 0 008 scan measured values consistently showed a standard deviation much higher than that value especially for roll The discrete and continuous inclination measurement methods were conducted separately from the scans such that the only movement during these measurements was the rotation of the scanner head The raw data of these methods is not available to the user as these proprietary methods simply provide roll and pitch values upon completion Although both methods produced results matching or improving upon the listed precision the continuous method gave the most consistent results despite the fact that the scanner head rotation speed during this method is similar to that of a 6 second overview scan Recall from Table 3 that the continuous meth
30. 0 and 27 after each iteration to prevent value inflation For each iteration a Jacobian matrix populated by the first order Taylor series polynomials of I re Yii M2 3 x Ty Y S 121 122 123 gt Ty th r31 132 133 Z T is calculated where T11 cos 8 cos 03 T12 cos 8 sin 0 sin 04 sin 0 cos 03 713 sin 04 sin 3 cos 04 sin 0 cos 03 To cos 0 sin 03 T 2 COS 04 cos 83 sin 04 sin 82 sin 03 sin 8 cos 63 cos 04 sin 0 sin 03 r31 sin 0 T32 sin 04 cos 0 T33 cos 04 cos 0 The K matrix as defined by Ghilani is then computed based on the results of the previous iteration or the initial guess for the first iteration and the corrections for the current iteration are computed by e JT WJ JTWK The inverse is computed using the C code provided by Ghilani 2010 in Table B 1 More efficient methods exist but this one works and is easy to implement Transformation parameters are then updated and the measured points are re transformed for the next iteration Once the exit conditions are met the results are printed to a text box Interpreting results Adjustment Results RMS 0 855544 Number of iterations 3 S 1 003665684 Tx 25463 973 Ty 25051 5655 Tz 20233 788 Rx 0 003904965 0 13 25 469 Ry 0 004509078 0 15 30 06 Rz 0 004864149 1 43 15 70 Rotation Matrik 0 999978 0 00486408 0 00450906 684825 425 bd4 754 367
31. 1AM Leica Faro custom 9 11 10 4 28AM Cyclone 9 7 10 2 08AM RiScan 9 6 10 1 21PM Leica Cyclone 6 9 4 10 6 11AM Leica Cyclone 7 9 4 10 4 25AM Propietary of Z F and LFM 9 3 10 11 44PM Cyclone Rapidform 9 3 10 10 48PM Leica Cyclone Register 9 3 10 2 35PM Leica Cyclone 7 9 3 10 3 15AM Question 8 Please share any comments or experiences you might have Count 17 When mixing 6100 and C10 data having the added constraint of a compensator in the C10 data greatly helps with overall scan cloud accuracy Registering the compensated data first then registering the un compensated 6100 data to the compensated data also aids in overall consistency with the data We treat compensated data as a being of a higher order of accuracy Inclination sensor information helps in many cases to reduce the ammount of needed targets Or in case to few targets are in sight it helps to 9 24 10 7 32PM register the scans at all Will you publish the results on the Laser Scanning Forum 9 24 10 3 17AM Never let an untrained person use a scanner you will never get the 9 23 10 6 29PM results are or they are expecting The ability to register scans on the fly provides a superior solution to cloud based regsitration The registered scans are viewable in the field 9 23 10 3 17AM and quickly identify areas of concern data gaps or descrepencies 9 26 10 5 32PM leica c10 seems to be fine with the compensator on all the time even in high wind Most of the work
32. 24PM Leica C10 9 22 10 7 02PM Leica SSII 9 22 10 6 14PM Leica c10 9 22 10 5 59PM Faro Photon 120 Leica Scan Station 2 9 22 10 4 21PM ILRIS 3D 9 22 10 10 33AM Riegl LMS Z620 Leica C10 Leica HDS6100 9 22 10 7 54AM Leica C10 9 22 10 5 35AM Leica HDS 3000 9 22 10 4 52AM Topcon GLS 1500 9 22 10 4 43AM Riegl VZ400 Leica 3000 9 22 10 4 39AM Leica C10 9 22 10 4 30AM Leica C10 laser scanner 9 22 10 2 42AM Leica 3000 and Cyrax 2500 9 22 10 12 54AM Leica C10 9 22 10 12 19AM Leica Scanstation HDS 3600 9 21 10 11 59PM Leica Scan Station C10 9 21 10 11 54PM HDS2500 9 21 10 10 52PM FARO LS880 9 21 10 10 26PM Leica HDS3000 9 21 10 7 55PM Leica SS2 9 21 10 6 06PM Leica C10 HDS6100 9 21 10 5 52PM Optech ILRIS 36D 9 21 10 5 31PM Leica C10 Scanstation 9 21 10 5 09PM All Leica HDS3000 ScanStation ScanStation 2 C10 and HDS6 00 9 21 10 4 26PM Leica C10 9 21 10 1 24PM leica scanstation 2 9 18 10 4 01AM Leica C 10 9 11 10 4 28AM Leica Scanstation 9 7 10 2 08AM Page 97 Riegl Z420i 9 6 10 1 21PM Leica Scanstation 2 9 4 10 6 11AM Leica ScanStation2 Leica ScanStation C10 9 4 10 4 25AM Z F 9 3 10 11 44PM Leica Scanstation2 C10 Surphaser and Riegl 9 3 10 10 48PM Leica ScanStation C10 9 3 10 2 35PM Leica Scansation C10 and Scanstation 2 9 3 10 3 15AM Question 2 Does your scanner have a level compensator Count 48 Yes 39 No 8 Not sure 1 What are level compensators 0 Question 3 If yes how often do you use your level compensator when scanning
33. AN ABSTRACT OF THE THESIS OF Evon P Silvia for the degree of Master of Science in Civil Engineering presented on June 1 2011 Title Overcoming the Level Bubble Terrestrial Laser Scanning Reference Frame Transformations Abstract approved Michael J Olsen Terrestrial Laser Scanning TLS is an efficient effective and precise measurement tool rapidly growing in popularity in a wide variety of fields Use of TLS data often requires aligning multiple scans for a more complete model of a scene or object a procedure known as scan registration Each scan setup must be transformed i e rotated scaled and translated and adjusted to match the other setups as closely as possible This research investigates and compares the traditional least squares method against a rapid noniterative technique recently proposed and develops possible solutions to achieve the strengths of both methods Additionally modern TLS instruments are commonly equipped with an inclination sensor to measure deviations from a level setup in order to simplify the registration process This thesis investigates the reliability usefulness and limitations of these sensors In particular the concept of a stability threshold that ensures reliable inclination measurements in modern rapid scanners is introduced and sources of inclination sensor error and procedures to minimize that error are investigated While level and unlevel setups appear to produce co
34. ANSFORMATION BETWEEN REFERENCE FRAMES Suppose a point Po in one reference frame is a vertical vector with coordinates x y and z Or Po 2 1 I NS X ee Converting a set of coordinates from one reference frame to another is known as transformation broken down into a three step process of rotation scaling and translation Page 9 These transformations can also be described using a quaternion which represents a single rotation about a vector Jekeli 2001 A quaternion can be computationally efficient and avoids the singularity problem known as gimbal lock and will be integrated into future research The approaches analyzed in this section do not use quaternions 2 1 1 Rotation The rotation of a reference frame around its origin can be described as a rigid body rotation using three angles known as roll or rotation around the X axis pitch or rotation around the Y axis and yaw 0z or rotation around the Z axis These angles are all measured relative to the original reference frame The rotation of a point s coordinates into a new reference frame using any one of those angles can be accomplished by multiplying Po by one of the following matrices 1 0 0 R 9 cos sin 2 2a 0 sin cos cos 0 sin y R 0 1 0 2 2b sin6y 0 cos0y cos sin 0 R sin cos o 2 2C 0 0 1 Page 10 The combined effect of Rx Ry and Rz can be summarized as Pp R R R Po RP 2
35. Others We also use the amberg GRP5000FX systems for mobile scanning in rail 9 26 10 5 32PM environments Registration in this sense is a chainage based process Combination of all above job dependent 9 23 10 6 29PM on the fly target registration for immeadiate registration reports and control 9 23 10 3 17AM confidence Traverse and backsighting from known point to known point 9 22 10 8 59PM Using the inclinator gives us a backup if some targets are missed 9 22 10 4 21PM We use occupy and back site as well as target registration which ever is 9 22 10 4 43AM better for the situation External survey control generated by a total station 9 21 10 11 54PM GPS coordinates for scan origin mostly used technique 9 21 10 5 31PM Cloud to Cloud 9 21 10 4 26PM A mixt of techniques depending on the results to be obtained 9 7 10 2 08AM I m a big fan of target based resections through known control Although requires preparation it combines mobility benefits of cloud surface 9 4 10 6 11AM matching with the some of targeted based registration modeled objects 9 3 10 10 48PM very rarely i have modelled surfaces or objects to use as registration 9 3 10 3 15AM Items Question 6 When registering your scans do you constrain the scan alignment to use the values from the level compensators select the most appropriate option Count 48 Yes 24 No 5 don t use a level compensator 8 It depends please explain 11 Others Not sure what you mean
36. The DriveVZ400 main screen Connecting to the scanner is fairly straightforward and can be done in one of two ways Wireless Once it has powered up the scanner becomes a wireless hotspot just like a normal router The network name is the same as its serial number S and uses WEP encoding Check with the lab manager to get the network password if you don t have it This connection can be a little slow when downloading raw data files Ethernet wired The VZ400 has two Ethernet ports one on the base requiring a special adapter 100MBit and one on the scanner head itself 1 GBit Most modern computers can switch cable type automatically but to be safe you should use a cross over Ethernet cable You can browse files on the scanner using Windows Explorer using either the serial number or the IP Address 10 0 0 1 wireless or 192 168 0 234 wired as the network address Note that DriveVZ400 defaults to using the scanner s serial number if it cannot connect and you know that you have connected try opening the IP address in Page 90 Windows Explorer first Sometimes the serial number does not register right away If your scanner has a different serial number you must edit the code For more detail on connecting to the scanner check the scanner owner s manual Using the program You only need to be connected to the scanner to use Fetch and Begin Scan Otherwise you do not have to maintain a connection to the scanner
37. The detailed analysis behind the development of that coordinate system was outside the scope of the manuscript presented in Chapter 3 This section will present discussion of the methodology and insights it provides on laser scanning technology The new coordinate system definition required four conditions see Figure 10 to simplify use in scientific research 1 The origin 0 0 0 must be located at the base and center of the wavemaker face as it rested on June 23 2010 2 The average of the floor defines the vertical datum Z 0 Page 58 3 The wavemaker face defines the Y axis such that X 0 along its length 4 The coordinate system must be right handed Note that north in the global sense follows the Y direction in the local coordinate system While this may seem counterintuitive such an orientation makes experimental results from the TWB consistent with the West Coast of the United States with waves approaching from the X direction or west Q O LL 2 O 4 gt z North Exterior Building Wall Scan Positions Benchmarks Targets Wavemaker Face Floor Cross Section 1 x Floor Cross Section 2 Floor Cross Section 3 Floor Cross Section 4 Figure 10 Plot of total station points recorded around the TWB in July 2010 Note that the wavemaker face is on the left in this figure with positive X pointing right and positive Y pointing up 4 1 1 Data Collection On J
38. Z400 exe m 1 n 20110222_test a 17 1080 41 p 42 134 43 570 02 321 897 456 123 02 i 0 r 1 The arguments follow a specific format It is important to note that there must be no spaces within a particular argument for the program to recognize it The actual order of the tags does not matter with the exception that the run mode m M must always come first Here is a rundown of each tag m M run mode This tag must come first o 0 Simply launch the program and ignore all other arguments o 1 Load the arguments into the program and wait for the user o 2 still to be implemented Load the arguments and start the scanner automatically n N project name This is the project name where all of the raw scan data will be stored usually on a USB drive plugged into the scanner Remember there can be NO SPACES even in folder or project names a A atmospheric conditions This is the air temperature C atmospheric pressure millibars and relative humidity of the current environment separated by commas o Temp range 50 to 100 C 1 precision 12 default o Pres range 600 to 1200mbar imbar precision 1000mbar default o RH range 0 to 100 1 precision 60 default Page 93 e p P scan preset pattern These are preset patterns with all of the parameters set up for you You ll probably default to the 2min for now This should be somewhat case insensitive o 2min 360 overview at 0 05 increments o 5min 360 o
39. a cc cecccccceeeseeeeeeee ees 7 2 1 Transformation Between Reference Frames ccccceccseeeeeeeeeeeees 8 2l ROAM ee ne eroreiaeonarioeoen 9 P RAE E 3 E EEA EE E EEA NE E ee ee ee EE 12 Zeo a aO E E E oaioansaee 13 2 1 4 Transformation SUMMALY serseriei arina E E E 13 22 POM Dalai AGUSIMEN E siie E E E 14 2 2 1 Least Squares Adjustment General Case cccceceeeeeeeees 15 2 2 2 Least Squares Adjustment Reference Frames 0008 17 2 2 3 NISLT A Least Squares Alternative ccccccecceecseeeeeeeeeees 18 224 NISLCTFOPIMZ ON cnnan a eid es 19 2 2 5 Adjustment Method Comparison c cccccccceccseeeeeeeeeeeeeeeeeees 20 2 3 Transformation and Adjustment ConcluSIOns cccccceeeeeeeeeeees 23 3 Man script Chapter seisnes a T 25 SW APSC anann N 26 Se INTOdGUCUON o E hl tout tcun tae 2 S2 Can IN COIS AtlOM irs corso shs eas atoe E S 28 322 JIMCIIN AON SEN SONS erpa A AEAN 30 3 2 9 OUIVeY Of TLS Professionals esicsscnciicseenstesaxesatscsaccatas ner aana 31 324 SCONE OF Manuscript eennnsnseniuanenennra nana aa aa 32 TABLE OF CONTENTS Continued Page 3 3 Inclination Sensor Accuracy StUCY ccccecceecceeeceeeseeeeeeeeeeeeeeseeees 34 3 3 1 Review of TLS Inclination Sensors cccccecceece eee eeeeeeeeeeeeeees 34 SZ PODEM OaE Il ena 36 3 3 3 Stability Threshold Evaluation cccceccecceseeceeeeeeeseeeeee
40. an arbitrary workable coordinate system 2 Translate in the Z direction such that on average Z 0 on the empty floor Page 60 3 Rotate around the Z axis such that the wavemaker face is parallel to the Y axis and the X axis points into the wave basin 4 Translate in the XY plane such that the new origin is in the center of the wavemaker face 5 Verify proper alignment between scans using the new control 6 Verify agreement between scan and total station data Repeat as necessary The end result of this work was a new coordinate system that has been used without any problems following its development It is worthwhile to note that the TLS data collection required significantly less field time than the total station and provided much more intuitive models with denser data Furthermore the coordinate system could have been developed using only one or two scan setups Three scan positions were used to provide redundancy to evaluate the quality of the scan data Page 61 Shift the total station data to fix NWBM to arbitrary coordinates Average the Y values along the 100m 100m 100m Use total north and south edge of the station target data as control wavemaker face 86 506750m y Using the X value of the wavemaker face origin and the centerpoint of the north and south edges of the wavemaker face adjust the control 144 105000m in the X direction and 86 506750m in the Y Isolate floor data from scans direction
41. an in Han Jen Yu Noniterative Approach for Solving the Indirect Problems of Linear Reference Frame Transformations ASCE Journal of Surveying Engineering 136 no 4 November 2010 150 156 The syntax for the command simply requires two windows style file paths to two comma delimited ASCII files one for control data and one for measured data A call to this function will look something like this adjustHan d mon Ctrl txt Js mon meas Cxt The function will match the points in order the first point with the first point the second with the second etc so ensure that they are sorted to match It uses the Matlab csvread filename function to import the points so follow the formatting rules for that function Output The function will output a series of text that will look something like the following gt gt adjustHan c temp 5pt ctrl txt c temp 5pt meas txt Raw Rotation Matrix R hat 0 407777391022836 0 913640618333182 0 000212293795449796 0 914294786828213 02407 352352919333 8 42174476038914e 005 0 0562338900691944 0 0297736416791388 0 99794047648357 Recomputed Rotation Matrix R calc 0 407569252666347 0 913174286505208 0 913174276498378 0 407569276741785 0 00021218543612765 8 43912511745261e 005 Q0 000163544184471777 0 000159356 7005009364 0 999999973927728 Derived rotation around x alpha rad deg 8 43912531744563e 005 Derived rotation around y 0 000212185437719842 Deriv
42. are platforms The compensator allows the individual to setup and level up over a known point and observe a single control point This is a far superior solution o photogrametric scanning solutions that require three or more control points The residuals for the control network are displayed immeadiately confirming that the cloud registration is within expected tolerances The dynamic compensator provides corrections during the data capture process which can be introduced by wind and or vibration The quality of the data with the compensator on verses off can be substantional 9 23 10 3 17AM Always use might not use when I take images some times but always 9 22 10 8 59PM when scanning 0 Use it to improve quality of the scans and be able to correct problems if 9 22 10 7 02PM they appear generally level all targets usig a total station anyway so sometimes dont 9 22 10 6 14PM use level compensator to speed up the process Allows easier cloud to cloud 2 picks Also i m a true surveyor and we like 9 22 10 5 59PM eveything to be levelled We have switched the compensator off on the scanstation to increase speed and in environments where the scanner needs to not be levelled scanning a ships deck However with the Faro we leave it on nearly all wee en the time Depends on control methodology as to whether use a the compensators Whenever set up directly over a mark the compensators are engaged 9 22 10 7 54AM Where suitable
43. ative although historical data indicate that the presence of this error resulted in the rotation and corruption of the wave basin s control network over several years as it propagated through subsequent surveys by the professional firm and lab technicians The use of the scanner s built in inclination sensor in this study provided a check on the control coordinates and quickly alerted operators to the problem demonstrating the value of using the inclination sensor as a quality control tool Additionally once new control coordinates had been determined for the wave basin Set 2 was then successfully registered to the local coordinate system using the same level constrained ICP software alignment as before saving operators the field time that would have been necessary to scan tiepoints for every scan position With the new control coordinates the alignment of Sets 1 and 2 were consistent as shown in Figure 9B and D The quality of the inclination sensor derived data was also confirmed using cross sections measured using a digital level and total station 3 5 CONCLUSION There is a tacit assumption among TLS users that inclination sensors in laser scanners behave similarly to their traditional operation in total stations For scanners that pause rotation during scanning this assumption appears to be sound as long as the scanner pauses for a sufficient amount of time It has Page 54 been shown that TLS inclination sensor results are
44. ays be the case with other instruments since grid scale factors and atmospheric corrections are generally assumed to be constant throughout the project site a similarity model is usually sufficient for laser scanning applications Affine models are more common in photogrammetry where directional dependent scale changes due to photograph distortion can occur Han 2010b 2 1 3 Translation Rigid translation of Pr is accomplished by adding a translation vector T to the point shifting it relative to the new origin Note that T is in the new reference frame its components would be different in the original reference frame if translation is applied prior to rotation 2 1 4 Transformation Summary In matrix form the full three dimensional transformation of point Po from one reference frame into another can be described with the following equation P SRP T X M1 M2 3 x ee Y S 1 122 23 gt 4 D 2 7 Z 131 132 133 Z T Page 14 Thus seven parameters are required for a three dimensional similarity transformation S 0x Oy Oz Ty Ty and Tz 2 2 POINT DATA ADJUSTMENT Every measurement contains error Lichti Gordon and Tipdecho 2005 Thus even if a given set of points were measured twice under identical conditions using a single instrument in perfect condition two slightly different sets of point data would be produced Assuming that only random error is present i e no mistakes o
45. cess is repeated until the corrections in e are sufficiently small Page 17 2 2 2 Least Squares Adjustment Reference Frames In the case of computing the seven parameters for a three dimensional similarity transformation that minimizes the overall error an iterative approach is required since 2 7 is nonlinear Each point used in the computation contributes three equations therefore a unique solution requires at minimum three non collinear points Using a traditional least squares adjustment four non coplanar points are required to prevent gimbal lock Jekeli 2001 Gimbal lock is a singularity condition in which a degree of freedom gets lost such that rotation around one axis has exactly the same effect as rotation around another axis This problem can also be avoided by adjusting pre leveled data simplifying the rotation matrix By analytically populating the matrices in 2 9 with 2 7 and its first order Taylor series approximation a standalone Qt based C program was developed to compute the seven transformation parameters required to match a set of measured data to a corresponding set of control data using the nonlinear least squares method Qt is an open source class based graphical user interface GUI extension for C See Appendix A for more details on this program and instructions regarding its use Page 18 2 2 3 NISLT A Least Squares Alternative The Nonlterative Solution to Linear Transformations NISLT
46. cility used for modeling coastal environments and simulating wave events and impacts Occupying a space approximately 49m x 27m x 2m 160ft x 87ft x 7ft the basin hosts large constructed scale models that must be properly recorded before and after testing Baseline scans and surveys of the basin itself were previously conducted by a 3D scanning firm to establish a local coordinate system anchored by three existing stable benchmarks on the facility floor Wall targets were surveyed see Figure 7 for a typical target scan and photograph to provide convenient tiepoints that could be used to orient and locate subsequent models DA TLEN r3 Mai i eo ee Figure 7 A typical HWRL target colored by return signal amplitude left and a grayscale photograph of the same target right Unfortunately this control proved to be unreliable and the lab manager reported difficulties properly closing surveys and linking data Page 50 3 4 2 Procedure For this case study scans were taken in the positions shown in Figure 8 distributed around the wave basin denoted by the shaded area focusing on the southwest corner of the wave basin which was the location of the scale building models being studied Precise leveling was not performed with each setup although inclination sensor data was recorded for each scan No new control or targets were placed for these scans because three to five targets were visible from each scan setup a sufficie
47. ction field time and eliminating the need to return to the site due to poor coverage A lightweight Qt based scanner interface was developed through this research in order to enable automatic collection registration adjustment and analysis of scan data from a Riegl VZ 400 See Appendix C for instructions in its use This program implements the inclination sensor recommendations developed in the previous chapters Page 68 Figure 16 Point cloud data set of a model beach at the HWRL after wave inundation Blue represents accretion and orange represents erosion nuns AWE Ss TE ee oes wae A ly OAA ini dij vnt EN eS ar Wat EER Re Figure 17 A gusset plate scanned before and after load testing colored according to deviations from a plane Red indicates large displacements Page 69 5 OVERALL CONCLUSION The research presented herein revealed current limitations of terrestrial laser scanning technology relevant to reference frame transformations and adjustments and recommended solutions to enable broader use 5 1 COORDINATE TRANSFORMATION AND ADJUSTMENT Least squares methods applied to a seven parameter three dimensional similarity adjustment traditionally require an iterative nonlinear solution This process can be computationally intensive and increasingly so as the data set increases in size This can be an issue when working with TLS data as millions of points can be collected in a single scan An alternative to th
48. d any speed related effects observed Nine 360 scans were collected from each setup at scan speeds ranging from 6 seconds to 740 seconds Past experience with the scanner suggested that the stability threshold was likely in the range of 60 to 400 seconds so the test emphasized speeds around that range which may vary between scanners Retro reflective targets with known coordinates were distributed throughout the room and scanned at the beginning of each setup to provide reliable external control for evaluating inclination sensor accuracy Coordinates for these control targets were defined using a carefully leveled Leica TCRP 1201 survey grade total station Page 38 In order to demonstrate that results were not peculiar to the inclination sensor in the scanner used for this study and that it was consistent with itself the discrete inclination measurement feature described as VZ 400 method 2 in Table 3 was used to record static inclination data after the scans were completed This discrete method resembles inclination sensor data collection from other scanner models that pause rotation while acquiring scan data Significant efforts were undertaken to reduce external errors sources during and between scans Solid connections between the scanner and tripod components restricted undesired rotation of the scanner Rather than using the keypad on the scanner head itself the instrument was operated remotely from a wireless laptop to e
49. d each individual axis R is the combined rotation matrix around the scan origin and P is a vector of the point coordinates in the new coordinate system Note that a translation along X Y and Z needs to be performed after rotation for complete registration When applying only level corrections Rz would be an identity matrix and y would be zero simplifying R Note that unlike airborne or mobile systems where a level correction is normally applied on a point per point basis through the use of an inertial motion unit IMU TLS level corrections are ordinarily applied to an entire setup with the assumption that the scanner platform is static Hence every scan and every point recorded at a particular setup receives the same correction 3 2 3 Survey of TLS Professionals The Oregon State University geomatics program conducted a brief online survey of laser scanning professionals in September 2010 see results in Appendix D The intent of the survey was not to be a rigorous scientific survey but rather to evaluate the industry s use and understanding of TLS Page 32 inclination sensors and applications 81 of the 48 respondents indicated that their scanners were equipped with an inclination sensor and 69 of the 48 respondents used it all or most of the time The survey results generally indicated an informed understanding of the scanner s inclination sensor and its uses Many TLS professionals who use inclination sensors stated
50. dation can be reduced cancelled or avoided through proper field techniques and scan planning A scan duration longer than a stability threshold ensures quality inclination data Alternatively measuring inclination data in both clockwise and counter clockwise directions separately from the actual scanning has also been shown to effectively eliminate inertial effects In all cases a full 360 rotation is recommended to remove systematic bias Page 71 5 3 TLS REFERENCE FRAME TRANSFORMATIONS IN THE FIELD A case study has been presented in which total station and TLS data were used to successfully develop a new coordinate system with a geometrically controlled definition The TLS data collection was significantly faster and provided a more robust solution for the geometric objects although the inclusion of some total station data also provided reliable checks against the TLS data when discrepancies were discovered Further study is required to determine the cause of these discrepancies TLS data collection also lends itself well to automation making it attractive for real time change detection applications A lightweight scanner interface has been developed to facilitate this automation through integration with other analysis tools Research in this area is ongoing at Oregon State University Page 72 WORKS CITED Agosto Eros Andrea Ajmar Piero Boccardo Fabio Giulio Tonolo and Andrea Lingua Crime Scene Reconstruction Usi
51. e traditional least squares solution presented in Han 2010b Known as the Nonlterative Solution to Linear Transformations or NISLT technique attempts to overcome this issue While the NISLT technique appears to be both fast and efficient the rotation matrix estimated by NISLT does not possess the properties of a true rotation matrix Therefore the derived rotation angles and the root mean square error from the NISLT adjustment may be unreliable It has been demonstrated numerically that a linear least squares solution or NISLT solution can be constrained using the properties of a rotation matrix in order to overcome this limitation but an analytical solution to implement this noniteratively remains to Page 0 be discovered Future research should investigate quaternion rotations as a possible solution Additionally the NISLT technique in its current form appears to be more sensitive to outliers than traditional least squares methods but more research using poor quality data is required to confirm this 5 2 INCLINATION SENSORS Inclination sensors in modern terrestrial laser scanners have been demonstrated to be useful and reliable when used properly For example inclination sensor data can be a valuable quality check on control data used for point cloud registration However it has been shown herein that inclination sensor data quality can degrade while scanning at high scanner head rotation speeds This inertial based degra
52. e 9 21 10 5 52PM erroneous readings LC lowers accuracy dual mirror system provides a very precise measurement routine in a internally fixed coordinate system once the scans are aligned to each another and geo referenced values from the Ic are redundant guess Ic s are needed for close range indoor scanning but not for long range out door activities as you will need several setups with precise gps coordinates for each scan origin 9 21 10 5 31PM We ll turn it off if the scanner has to be used while non level Boats ships 9 21 10 5 09PM While tilted to achieve more down angle not use when on a boat or a moving object when the scanner is moving at the same rate as the object Use all the time the SS2 and C10 are survey 9 21 10 4 26PM instruments The compensator gives you a virtual target many kms up in the air 9 21 10 1 24PM On a project where we used it we had a great deal of difficulty getting good results with the registration Using 4 targets and the resection 9 18 10 4 01AM method gives us much better results Allows individual scans to be controlled using fewer unknowns during 9 11 10 4 28AM registration To obtain at least one scan with real verticality of the site 9 7 10 2 08AM Aids in the accuracy when using a single target backsight method 9 6 10 1 21PM We use our laser scanner mostly for topographic surveys Under changing weather conditions or unstable surfaces e g hot asphalt the weight and movement of the
53. e is detected Measurement Program High Speed works for most applications Using Long Range increases scan time by approximately 2 5 times but can increase measurement range Atmospheric conditions lf you brought a thermometer and barometer with you enter in the atmospheric conditions and the scanner will automatically correct measured values Temperature is in degrees Celsius barometric pressure is in millibars and relative humidity is in Page 91 Relative humidity has a minimal effect 0 25ppm if you default to 50 so do not worry too much about it if you are unable to measure that Scan settings This is the meat of the program All angles are measured in degrees and define the scan window to be recorded by the scanner Vertical angles are measured from the vertical axis going down and horizontal angles are measured counter clockwise from the scanner s power port up to 720 See the next section for details on increment precision Usually you need to do no finer than thousandths of a degree While running a scan verify that the parameters received by the scanner are the same as the ones entered Common presets are provided based on scan time add more by editing the VZConfig cpp code file The Default scan does a scan at the coarsest possible resolution 2min and 5min are good for medium and high density 360 scans To get an estimate of how long a scan with the given parameters should take click
54. ed primarily by Besl and McKay 1992 has seen many developments in recent years including in the context of laser scanning This software based technique uses common surface Page 29 features possessing distinct geometry to match scans to a reference scan without the use of targets For a summary of modern applications of and improvements to ICP see Vosselman and Maas 2010 and Salvi et al 2007 e Attempting to scan large geographic areas using TLS presents unique challenges including the need to quickly georeference data to a real world coordinate system A surface matching azimuth adjustment technique Olsen et al 2009 2011 has been shown to produce repeatable results within equipment specifications particularly over long coastal sections This technique determines scanner origin coordinates using RIK GPS constrains scans to level using inclination sensors and conducts a software adjustment to the azimuth using a point to plane ICP variant based on matching features common to multiple adjacent scans some of these techniques require the field data to be close to level for reliable and efficient operation Field procedures and time constraints can inhibit exact instrument leveling at every setup and so many TLS manufacturers have provided means to compensate for unlevel setups both in the instrument itself and in accompanying software As dependence upon proper compensation increases so must the scrutiny of these device
55. ed rotation around z 1 15100570648747 Translation vector 34788 213974199 26069 4690880224 0 00483526263471635 beta rad deg 0 012157330055483 gamma rad deg 65 9477691772056 73 8516663028582 Updated translation vector 34788 213499084 26069 4773608414 73 8513886331855 Scale 0 999663999733704 Transformed points R hat 34758 4534357 34775 3889262245 34795 3742079115 34819 1617829325 34775 95326472315 Or R calc 34758 4513983188 34775 4015314734 34795 3713596344 34819 1680899039 34775 5186206694 RSME R hat R Calc new T 0 00573323275059479 0 01572208177502352 Optimizations One simple optimization was made to the original NISLT method to reduce memory usage and computation time Rather than computing scale and the AX terms over the range of j we computed for j gt i where i increments from 1 to n 1 and j increments from ito n where nis the number of point pairs This removes redundancy introduced by computing the same distance ij and j twice 26096 26L05 26087 26061 26077 26096 9804999565 9926083108 4272506351 7720804411 04956006564 9688260255 26105 987775405 26087 427098077 26061 26077 7938219339 0444785586 Page 86 5920597106532 0473863268673 8420325614988 0923313997251 3751900012557 5933029359046 0474972076757 8417244584684 0904835942779 3759918036734 The pseudo rotation matrix R hat R Scale the Tran
56. eeees 37 3 3 4 Rotation Speed Effects eE 41 300 OCannenPosilion ENGCIS airna a enna 44 3 3 6 Scanner Directional Effects cc eeccccseeceeeeeeeeteeeeeeeeeeeeeees 45 3 4 Case Study O H Hinsdale Wave Research Laboratory 48 Ore BACK OROUING scarsia E asennad aioanaaiaeinenmasonenmaianaumanceiaats 49 DA 016 26 0 fs ee nn en ear or ce ee ee ce eee hr Te 50 ero ANIV SIS eee ee ee ee ee eee 50 34A RESUS eeri A bamieoeanamiaaaamiaaennaes 52 J9 CONCIS ION we tvadctcntunsetontmonssoninonssanhoonasondsonsesinods tesisosatonsonadinnta tania 53 96 FACKMOWIGCOCINICING etait aunts saeco G and sent aet an tanbs abahavat sean aac erates 56 4 Field Applications of TLS Reference Frame Transformations 57 4 1 The HWRL Tsunami Wave Basin Coordinate System 08 57 AAA Data CONC CUON wvarevsssvetasnvsvatanntovedineldvalavertenteuaiausiavebowolavavavabeunbes 58 4 1 2 Development of the New TWB Control Coordinate System 59 4 1 3 Additional Considerations ccccccecceeceeeeeeceeeeeeseeeeeeeeeeeeeeaes 62 4 2 Real time Change Detection cc cccceccceccceeceeeeseeeeeeeeeeeeeeseeeeeeess 67 5 Overall CONGCIUSION erisera E EEAO 69 5 1 Coordinate Transformation and Adjustment cccceceseeeeeeeeeees 69 32 IMCIIMANON SEN SONS ososrosen ninna nacre 70 5 3 TLS Reference Frame Transformations in the Field 06 71 WorkS Cited
57. es not demonstrate Property 3 of a rotation matrix Section 2 1 1 and is therefore not a true rotation matrix This is likely because the NISLT technique solves for the nine elements of the rotation matrix as nine independent variables when in fact a rotation matrix consists of nine dependent variables calculated from three independent angles roll pitch and yaw This increased number of degrees of freedom may explain why the RMSE using R appears to be lower than what would be theoretically possible The discrepancy between R and R calc has two very important implications Firstly a user of the NISLT technique cannot rely on the extracted roll pitch and raw as implied by Han 2010b but should use R directly in all subsequent computations Secondly the NISLT in its present form appears to be more Page 24 sensitive to outliers than traditional least squares methods but more research using poor quality data is recommended to further evaluate this While the R mod solution was very slow compared to R LS and R its low RMSE and rotation angles comparable to R LS demonstrate that an analytical application of one constraint to the NISLT technique could theoretically maintain the advantages of the NISLT technique while overcoming its shortcomings To the author s knowledge no such analytical solution has yet been found Page 25 3 MANUSCRIPT CHAPTER To Level or Not to Level Laser scanner inclination sensor accuracy evaluat
58. f vz gt waitForFinished 1 this gt userUpdate Program finished this gt userUpdate QString 1 arg vz gt exitCode delete vz QProcess start runs the child program in a separate thread so the host parent program will still be running at the same time under another process The OProcess waitForStarted 1 and OProcess waitForFinished 1 functions lock the parent program until an exit signal is received from the child program These functions are not required but they can be helpful if your parent program cannot continue without the scan information Page 95 APPENDIX D RESULTS OF A SURVEY OF TLS PROFESSIONALS Page 96 Industry survey of TLS Professionals This is a detailed summary of a survey conducted in September through lidarnews com 13 responses and in October through laserforums uk 85 responses Text based responses are time stamped in order to provide cross referencing Question 1 What is the make and model of your laser scanner s Count 48 Leica HDS Scan Station 2 9 28 10 12 53AM Leica C10 Leica 6100 9 26 10 5 32PM Trimble GS 101 9 24 10 9 14PM FARO 9 24 10 7 32PM HDS 6000 ScanStation LS880 9 24 10 4 07AM HDS3000 9 24 10 3 17AM faro Is 120 9 24 10 1 47AM Riegl 390i 9 24 10 1 25AM ScanStation C10 9 23 10 6 29PM have worked exclusively with Trimble scanners mainly the GXA 9 23 10 3 17AM Leica ScanStation 9 22 10 8 59PM Trimble FX 9 22 10 8 08PM FARO Photon 120 9 22 10 7
59. htly out of level is good enough NEIDA PORN Very rarely 9 3 10 10 48PM Question 7 What software package s do you use for scan alignment Count 48 Cyclone 9 28 10 12 53AM cyclone latest version Amberg TMS 9 26 10 5 32PM Trimble Realworks 9 24 10 9 14PM SCENE 9 24 10 7 32PM Cyclone 9 24 10 4 07AM Cyclone v6 9 24 10 3 17AM Faro Scene 9 24 10 1 47AM Riscan Cyclone 9 24 10 1 25AM Cyclone Register 9 23 10 6 29PM The alignment is performed in the field Trimble Pointscape or Trimble 9 23 10 3 17AM Access Leica Cyclone 9 22 10 8 59PM Trimble Realworks 9 22 10 8 08PM FARO Scene 4 7 9 22 10 7 24PM Leica Cyclone 9 22 10 7 02PM Cyclone 9 22 10 6 14PM Cyclone pointools geomagic 9 22 10 5 59PM Faro Scene Cyclone 7 2 9 22 10 4 21PM Polyworks 9 22 10 10 33AM Riscan Cyclone 9 22 10 7 54AM Cyclone register 9 22 10 5 35AM Leica Cyclone Leica Cloudworx software developed by my company 9 22 10 4 52AM Topcon s Scan Master Software 9 22 10 4 43AM RiScan Pro or Cyclone 9 22 10 4 39AM Leica Cyclone 9 22 10 4 30AM Leica Cyclone 9 22 10 2 42AM Page 102 Cylcone 9 22 10 12 54AM Leica Cyclone 9 22 10 12 19AM Cyclone 9 21 10 11 59PM Cyclone 9 21 10 11 54PM Leica Cyclone REGISTER 9 21 10 10 52PM FARO Scene 9 21 10 10 26PM Cyclone 9 21 10 7 55PM Cyclone 9 21 10 6 06PM Cyclone 9 21 10 5 52PM JRC 3D Reconstructor 2 PolyWorks 9 21 10 5 31PM Cyclone 9 21 10 5 09PM Cyclone 9 21 10 4 26PM Leica Cyclone 9 21 10 1 24PM Leica Cyclone 7 0 9 18 10 4 0
60. in a scan window from 120 to 180 To verify this average roll and pitch inclination values were computed from 60 subsets of the 3 minute scan from the previous section Results for each interval varied from the overall average by up to 0 02 for both roll and pitch This difference can be significant as the difference between 0 02 and 0 02 translates to a vertical offset of approximately 3 5 cm at 50 m It should also be noted that there was some variability observed in initial measurements due to scanner park position which is the rotating scanner head s resting position relative to the position where it begins measurements These effects were removed in the scans shown by making the park position and start position identical Page 45 3 3 6 Scanner Directional Effects Now that it has been shown that the inertial effects observed in inclination data recorded during scan rotation increase with rotation speed it is possible that this error can be partially or completely mitigated by recording inclination data in both clockwise and counter clockwise directions If the inertial error is systematic it should theoretically cancel out if inclination data is recorded in both rotational directions 3 3 6 1 Scanner directional effects assessment methodology Twenty 6 second 360 overview scans were recorded from a moderately unlevel setup alternating clockwise and counter clockwise such that ten were recorded in a clockwise di
61. in the following line will always remain the same There will always be a blank line after each tag set e VZ_ PROJ the following line will be a string with the project s name on the scanner s storage device e VZ PATH the following line will be a string with the path to the downloaded monitor RXP file e VZ_INCL the following two lines will be floats with the roll and pitch values for the current instrument setup Calling another program using Qt There s a couple ways to run one program from within another QDesktopServices openURL is a very useful function that opens standard files like txt or doc files using the default program registered with the operating system but it doesn t really work for the DriveVZ400 program since there s no file type associated with it Page 94 QProcess is a class in Qt intended for running programs in another thread Here s some sample code that calls the DriveVZ400 program using the arguments described above QProcess vz new QProcess this Q StringList args args lt lt m 1 lt lt n 20110222 test23 lt lt a 19 1043 70 lt lt p 395 457 200 155 Les 1 lt lt i 1 lt lt r 1 vz gt start C programs DriveVZ400 exe args if vz gt waitForStarted 1 this gt userUpdate Program started qApp gt processEvents execute any events collected in the mainEvent queue such as displaying the line I just output before executing the child program i
62. influenced by scanner motion during scanning and therefore must be treated differently than those of total stations particularly for rapid scanners While some of the inertial error can be reduced by measuring inclination data during scans in both a clockwise and counter clockwise direction some random error will still exist possibly due to the physical mechanics of measuring during laser scanning This inertial error also varies with the rotating scanner head s position at the time of measurement therefore scans used to measure precise inclination data should encompass a full 360 horizontal scan window Even on those occasions when a scanner is leveled as accurately as possible a small amount of error will always exist Unless inclination data is recorded for each setup it is difficult even impossible to determine by inspection whether a point cloud is actually leveled particularly since the scan setup is generally close to level A properly leveled point cloud is significantly easier to work with and transfer between software packages than one tilted in space Leveling tests were conducted in a controlled lab setting with the inclination sensor enabled These tests investigated whether scanner rotation speed correlated with inclination data quality and then developed a recommended minimum 360 scan duration a stability threshold that ensured reliable Page 55 inclination data For the scanner used in this study
63. ion Evon P Silvia and Michael J Olsen 1 Graduate Assistant School of Civil and Construction Engineering Oregon State University 220 Owen Hall Corvallis OR 97331 evon silvia gmail com 2 M ASCE Assistant Professor School of Civil and Construction Engineering Oregon State University 220 Owen Hall Corvallis OR 97331 michael olsen oregonstate edu ASCE Journal of Surveying Engineering http ascelibrary org suo Manuscript under review Submitted on April 12 2011 Manuscript FSUENG 191 Page 26 3 1 ABSTRACT Many modern terrestrial laser scanners TLSs are equipped with inclination tilt sensors also referred to as level compensators that can correct out of level imperfections in an instrument setup Some users elect to disable the inclination sensor when their scanner is equipped with one Those who do use an inclination sensor may override its leveling data by linking to external control that defines the level plane Our case study has shown that inclination sensor data can be a valuable quality check on the control data and overall point cloud alignment Rigorous testing revealed that caution must be exercised in scan planning to ensure quality inclination sensor data Specifically lab tests indicate that scanner rotation speed influences the reliability of inclination sensor readings recorded during rotation These inertial effects can be cancelled to a limited extent by measuring during both clockwise and
64. k D and Maria Gabriele Licht Experiences with Terrestrial Laser scanner Modelling and Accuracy Assessment International Archives of Photogrammetry and Remote Sensing 36 no Part 5 September 2006 155 160 Lichti Derek D Stuart J Gordon and Taravudh Tipdecho Error Models and Propagation in Directly Georeferenced Terrestrial Laser Scanner Networks Page 75 Journal of Surveying Engineering ASCE 131 no 4 November 2005 135 142 Mechelke Klaus Thomas P Kersten and Maren Lindstaedt Comparative Investigations into the Accuracy Behaviour of the New Generation of Terrestrial Laser Scanning Systems Edited by A Gruen and H Kahmen Optical 3 D Measurement Techniques VIII Zurich 2007 319 327 Olsen Michael J Elizabeth Johnstone Falko Kuester Neal Driscoll and Scott A Ashford New automated point cloud alignment for ground based light detection and ranging data of long coastal sections Journal of Surveying Engineering ASCE 137 no 1 February 2011 14 25 Olsen Michael J Elizabeth Johnstone Neal Driscoll Scott A Ashford and Falko Kuester Terrestrial Laser Scanning of Extended Cliff Sections in Dynamic Environments Parameter Analysis Journal of Surveying Engineering ASCE 134 no 4 November 2009 161 169 Ravani Bahram Jagannath Hiremagalur Kin S Yen Kevin Akin Triet Bui and Ty A Lasky Creating Standards and Specifications for the Use of Laser Scanning in Caltrans Projects
65. l Prof Schultz Dr Mike Olsen deserves my sincerest appreciation for taking me on as a grad student within half an hour of our first meeting for expediting the paperwork to get me into grad school for persevering whenever got difficult and for always finding enough funding to keep me going Your patience understanding flexibility and instruction made the grad school experience rewarding even if didn t pursue a PhD like you wanted Thanks to Amber Clark and Dr Tom Dick for your thought provoking help with the mathematics Finally a big thanks goes to my brother and sisters who stayed up late with me while worked on my papers homework and projects and who tolerated me when got inevitably sleep deprived and loopy CONTRIBUTION OF AUTHORS Dr Michael Olsen assisted with the interpretation of the mathematics and data in Chapter 2 He also contributed to the experimental design editing analysis and direction of Chapter 3 Tim Maddux provided the historical data and access to the O H Hinsdale Wave Research Laboratory presented in Chapters 3 and 4 Shawn Butcher assisted with data collection TABLE OF CONTENTS Page E UAMOOGUGCUOMN its tisoiaioticaisiolioninretiiaiaiwiieeaiaioveoaiadanerpiadonwsaiedaccrolainieimiciquselaimens 1 Tel cLaSeF SCanning OVESIVISW ecucdeececreteriaresdvncetncticcesuasseretostensadortieudeetaeeets 1 t2 THESS FOMA conreen a o nE 5 2 Transformation and Adjustment of Point Dat
66. le prior to scan acquisition Precision unlisted Riegl VZ 400 1 Continuous during scan automatic Range 10 other models 2 Discrete 360 rotation pausing Resolution 3 6 similar every 60 for 1 second to measure Precision 0 5 semi automatic 3 Continuous both clockwise and counter clockwise Semi automatic TLS specifications should include some statement regarding the precision of its built in inclination sensor the Riegl VZ 400 used in this study lists its one sigma value at 0 008 degrees or 28 8 seconds Like most laser scanner specifications the exact procedure used to determine these values appears to vary from manufacturer to manufacturer and comparing scanners can be difficult though some work has been done to develop a common basis for comparison Mechelke Kersten and Lindstaedt 2007 Page 36 3 3 2 Problem Statement From the comments given in the industry survey described previously there appears to be a tacit assumption in the industry that inclination sensors in laser scanners behave similarly to those found in total stations However to reduce scan durations many new scanners simultaneously acquire point data and level corrections That is unlike total station operation the inclination sensors in these newer scanners are not static while measurements are taken In the case of a rotating scanner head if the inclination sensor is not located precisely in the center of rotatio
67. liable 3 3 4 1 Rotation speed comparison methodology Full 360 scans of durations of approximately 0 1 minutes 0 5 minutes 1 minute 3 minutes 5 minutes and 10 minutes were recorded in a controlled lab setting from a single stable unlevel scan setup As observed in the Page 42 Stability threshold evaluation the same effects were observed regardless of whether or not the setup was carefully leveled Every scan was recorded in the clockwise direction These particular scan durations were selected because they straddled the stability threshold determined in the previous section Thus measurements from scans above the stability threshold should show statistically the same variability while measurements from scans below the stability threshold should show higher variability Individual inclination measurements were extracted from each scan s raw measurement data As before roll and pitch described throughout this section are shown relative to the fixed scanner reference frame not the scanner head orientation at the time of measurement 3 3 4 2 Rotation speed comparison results and discussion Figure 5 shows radial plots of roll top and pitch bottom from four of the six scans recorded The two longer scans showed the same tendencies as the 3 minute scan and were excluded from the plots for clarity A perfect circle with minimal noise would indicate a constant and extremely stable inclination measurement The inclinatio
68. liminate any disturbances introduced by pressing buttons on the scanner head and having cables hanging from the scanner Once setup was completed no contact was made with the scanner head for the duration of each setup 3 3 3 2 Stability threshold results and discussion The results of the evaluation are presented in Table 4 and Figure 4 The data show a significant shift between the 30 second scan and the 6 second scan in the Leveled setup that skews the data This extreme variability of inclination values from the 6 second scan had been observed during previous scans so rather than removing it as an outlier it seemed fitting to retain the data to Page 39 demonstrate the risk associated with using inclination sensor readings obtained during extremely fast scans The inclination values computed from a registration to external control is provided in the tiepoint TP registration row Discrete method DM and software SW residuals were computed against the TP average and found to converge toward zero as the scan duration increased Table 4 Test 2 results for 360 scans from both unlevel and level scans DM roll and pitch values measured using the discrete method described in Table 3 SW roll and pitch values extracted from the raw scan data using the scanner s accompanying software NOT LEVELED Time Elapsed sec o SS DM res SW SWres DM DM res SW SWres _ 740 0 02 _ 0 723 45 0o25 0714
69. lution of 2 11 constrained by the six conditions in 2 6 in an attempt to force the properties of a rotation matrix onto R The six constraints counteracted the six additional unknowns introduced by solving for the nine elements of R instead of the three rotation angles granting the numerical solution the same degree of freedom as a traditional least squares solution Page 22 Table 1 Comparison of 3x3 rotation matrices from different adjustment methods for a set of GPS points with and without outliers Italicized values are the sum of the squares of the corresponding row or column 10 Point Solution with outliers 8 Point Solution without outliers R RMSE 0 1117 R RMSE 0 0196 0 40628 0 91328 0 03713 1 00052 0 40820 0 91438 0 06065 1 00639 0 90952 0 41107 0 16295 1 02275 0 91335 0 40780 0 00846 3 1 00058 0 01528 0 01458 0 31532 0 09987 0 00037 0 00125 0 94026 0 88409 0 99252 1 00327 0 12736 1 00084 1 00238 0 88784 R calc RMSE 1 0565 R calc RMSE 0 0416 0 40781 0 91178 0 04843 3 1 00000 0 40803 0 91297 0 00137 1 00000 0 91294 0 40807 0 00486 1 00000 0 91297 0 40803 0 00021 1 00000 0 01534 0 04620 0 99882 1 00000 0 00037 0 00133 1 00000 1 00000 1 00000 1 00000 1 00000 1 00000 1 00000 1 00000 R mod RMSE 0 1304 R mod RMSE 0 0233 0 40728 0 91330 0 00408 1 00000 0 40791 0 91302 0 00023 1 00000 0 91329 0 40730 0 00350 1 00000 0 91302 0 40791 0 00067 3 1 00000
70. methods are usually processed in the office although research to interact with the data in real time is ongoing Skaloud et al 2010 1 2 THESIS FORMAT This document follows the manuscript thesis format Chapter 2 introduces the basic principles of transforming point data between reference frames The adjustment of TLS related point data will also be discussed covering both traditional least squares adjustment methods and a noniterative approach developed by Han 2010b The shortcomings and effectiveness of each method will be discussed Chapter 3 presents a technical manuscript currently under review for publication in the ASCE Journal of Surveying Engineering on applications and assessments of inclination sensors in terrestrial laser scanning The Page 6 manuscript also includes a case study demonstrating the value of inclination sensors for discovering control network errors Chapter 4 extends the discussion of the new O H Hinsdale Wave Research Laboratory control coordinate system introduced in Chapter 3 to summarize considerations that arose in its development Also presented will be an application of TLS reference frame transformations for real time change detection including a custom interface developed to expand the capabilities of the scanner used in this research Chapter 5 summarizes the conclusions and contributions of this thesis and presents further study made possible by this research The appendices c
71. n a near impossible feat given that many scanners can be set up out of level then it would follow intuitively that while inclination sensors can stabilize during slow scans inclination sensors would be unable to reach vertical equilibrium during fast scans That stability threshold where the inclination sensor achieves near equilibrium prior to measurement must be determined for each scanner The following sections describe an investigation into the existence of such a threshold and other factors contributing to inclination sensor error Please note that roll and pitch described throughout this section are measured relative to the fixed scanner reference frame not the scanner head orientation at the time of measurement Page 3 3 3 3 Stability Threshold Evaluation The purpose of this evaluation was two fold 1 to verify a correlation between scanner rotation speed and inclination sensor stability and 2 to develop a methodology to determine a minimum recommended scan duration that ensures the stability of the inclination sensor that is a stability threshold for the inclination sensor This test was intended to be straightforward enough to determine the stability threshold of any similar scanner 3 3 3 1 Stability threshold evaluation methodology The evaluation involved two independent setups one carefully leveled and the other intentionally out of level to determine whether the magnitude of inclination measurements amplifie
72. n sensor used in this study was very sensitive to environmental disturbances including nearby slammed doors and passing trucks which is evident in the sudden spike around 190 for the 3 minute plot Page 43 315 13 45 m A M 0 9 Roll value of 2 510 at 0 8 72 horizontal lies outside of this plot 0 7 y p A amp y o R Legend mmm Ro for 3min Scan Roll for 1min Scan Roll for 0 5min Scan Roll for 0 1min Scan Legend meme Pitch for 3min Scan Pitch for 1min Scan Pitch for 0 5min Scan Pitch for 0 1min Scan Figure 5 Radial plots of roll top and pitch bottom measurements radial axis in degrees from level at approximate horizontal scanner head rotation positions polar axis Scan durations listed are approximate Measurements were taken approximately once per second Page 44 As expected individual measurements from scans with durations shorter than this scanner s stability threshold of 100 seconds generally show higher variability from a perfect circle than longer scans Figure 5 This indicates a direct correlation between scan rotation speed and error amplitude for an individual measurement which is confirmed by a computation of the standard deviations of each scan 3 3 5 Scanner Position Effects Intuitively one would assume that inclination data from a 360 scan would be more representative of the overall scan setup than a scan that for example only recorded
73. ng a Fully Geomatic Approach Sensors 8 no 10 2008 6280 6302 Anderson James M and Edward M Mikhail Surveying Theory and Practice 7th Edition WCB McGraw Hill 1998 Barber David Jon Mills and Sarah Smith Voysey Geometric validation of a ground based mobile laser scanning system SPRS Journal of Photogrammetry and Remote Sensing 63 no 1 January 2008 128 141 Bernardini Fausto Holly Rushmeier loana M Martin Joshua Mittleman and Gabriel Taubin Building a Digital Model of Michelangelo s Florentine Pieta IEEE Computer Graphics and Applications 22 no 1 Jan Feb 2002 59 67 Besl P J and N D McKay A method for registration of 3D shapes IEEE PAMI 14 no 2 1992 239 256 Boehler W M Bordas Vicent and A Marbs Investigating Laser Scanner Accuracy Updated Online 2004 i3mainz April 2004 http scanning fh mainz de scannertest results200404 pdf accessed November 15 2010 Page 73 Cuartero Aurora Julia Armesto Pablo G Rodriguez and Pedro Arias Error Analysis of Terrestrial Laser Scanning Data by Means of Spherical Statistics and 3D Graphs Sensors MDPI November 2010 10128 10145 Dunning S A N J Rosser and C I Massey The integration of terrestrial laser scanning and numerical modelling in landslide investigations Quarterly Journal of Engineering Geology and Hydrogeology 43 2010 233 247 FARO Technologies Inc FARO LS 880 Techsheet English January
74. ngle Lichti Gordon and Tipdecho 2005 Efforts to produce standardized accuracy tests for TLS are ongoing in multiple contexts including ISO standards Cuartero et al 2010 manufacturer independent instrument comparisons Mechelke Kersten and Lindstaedt 2007 cultural heritage scanning Boehler Vicent and Marbs Page 28 2004 construction standards Ravani et al 2007 and self calibration Lichti 2007 Lichti and Licht 2006 3 2 1 Scan Registration Further advancement of TLS technology in recent years has included improvements in scan alignment techniques Registration of scans to targets setup at known or externally measured positions is a common method for TLS alignment However safety and time constraints can make target registration impractical leading to the development of other registration techniques Three example situations and solutions are e Construction and underground mining sites operate on tight schedules that must minimize interruptions and cost Operation of dual axis inclination sensors enables faster setup by requiring fewer targets for scan registration Ravani et al 2007 reducing the cost of replacing targets disturbed during the construction process and reducing the safety risks associated with target placement without losing accuracy e Cliff and landslide monitoring applications make the use of targets a safety risk to field crews The iterative closest point technique ICP originally develop
75. nsistent results full 360 scan windows measured in both the clockwise and counter clockwise direction are recommended Finally some field applications of TLS reference frame transformations are presented and discussed including the development of a coordinate system in a large scale research laboratory and steps toward the development of real time change detection Copyright by Evon P Silvia June 1 2011 All Rights Reserved Overcoming the Level Bubble Terrestrial Laser Scanning Reference Frame Transformations by Evon P Silvia A THESIS submitted to Oregon State University in partial fulfillment of the requirements for the degree of Master of Science Presented June 1 2011 Commencement June 2011 Master of Science thesis of Evon P Silvia presented on June 1 2011 APPROVED Major Professor representing Civil Engineering Head of the School of Civil and Construction Engineering Dean of the Graduate School understand that my thesis will become part of the permanent collection of Oregon State University libraries My signature below authorizes release of my thesis to any reader upon request Evon P Silvia Author ACKNOWLEDGEMENTS would like to thank my committee members Michael Wing Mike Bailey and especially Robert Schultz for their patience and time and for their assistance reviewing this thesis could not have started working toward my Masters without the support of the immorta
76. nslation vector can then be estimated by calculating the average shift between the rotated scaled points and the transformed points t mean x Rx 2 13 An implementation of this method using a MATLAB script is presented in Appendix B with instructions for its use 2 2 4 NISLT Optimization A simple modification can improve the computational efficiency of the NISLT similarity model This optimization was implemented in the MATLAB script and discussed in this section As defined in the previous section dx will be equal to dx when formulating the matrices similarly dx will be equal to dx Since all terms are equally weighted computing both dx and dx and dx and dx unnecessarily doubles the computations and memory usage in 2 10 and 2 12 Therefore Page 20 these equations can be modified by changing i j to f gt where i increments from 1 to n 1 and j increments from to n where n is the number of point pairs This also reduces the number of rows in the matrices AX and AX by half from n n 7 to n n 7 2 reducing the required memory to store the matrices and subsequently the computations required for 2 11 For example in one test run with 19 point pairs this optimization reduced the dimensions of AX and AX from 342x3 to 1 713 noticeably reducing computation time without altering the final results 2 2 5 Adjustment Method Comparison While computation time for
77. nt number for registration Ke mM _ 2 09 gt lt LLI Figure 8 The HWRL tsunami wave basin building walls are outlined the wave basin itself is shaded Scan positions are denoted SP with an arrow indicating setup orientation Brass floor benchmarks are denoted BM 3 4 3 Analysis Scan positions 1 3 5 7 and 8 Set 1 were registered to the existing local coordinate system using the established targets along the lab walls To reduce Page 51 field time and because the dataset was also going to be used as a training dataset for processing it was intended to register scan positions 2 4 and 6 Set 2 to Set 1 using a level constrained ICP software alignment While the standard deviations were acceptable on the order of 2 4 mm for the target registration of Set 1 surface matching of Set 2 to Set 1 proved unsatisfactory Figure 9 Representative iso views and cross sections 40x vertical exaggeration of the wave lab roof after surface registration of scan position 2 light red to scan position 7 dark black Scan position 2 is approximately 7cm below scan position 7 on the west end and 7cm above it on the east end as shown in A and C Once the control had been corrected the scans aligned properly as shown in B and D Further investigation revealed a consistent directional bias as demonstrated in Figure 9A and C indicating a problem with the control rather than a problem
78. nt value as a quality control and scan registration tool While manufacturers often list precision specifications for a sensor in a particular instrument these sensors occasionally show more variability in actual operation than during strictly controlled testing This led to an investigation into one probable cause inertial disturbance from scanner rotation during readings 3 3 1 Review of TLS Inclination Sensors How each scanner measures inclination sensor data varies with manufacturer and data storage and extraction occur within a black box from the end user s perspective Table 3 summarizes inclination sensors for some modern scanners Page 35 Table 3 Summary of inclination sensor measurement techniques of four example scanners Modified from JADA Productions 2010 FARO Technologies Inc 2009 and Leica Geosystems AG 2011 Inclination Sensor Range Scanner TOE from horizontal plane Inclination Sensor Measurement ae Model Resolution and Precision 1 sigma FARO LS 880 Continuous after scan while recording Range 15 photographs Resolution 3 0 Precision 6 3 Leica Incremental pausing to measure for Range 0 083 ScanStation 2 each vertical scan line If significant Resolution 1 and C10 changes are detected the scan is Precision 1 5 aborted automatic Maptek I Site Performs leveling analysis during a Range 5 8800 continuous 1 minute 360 preview Resolution 20 cyc
79. od like the paired scan method measures in both clockwise and counter clockwise directions if the degradation of inclination data was from scanner rotation speed alone the two methods should have produced similar results This suggests that another factor introduces random error during laser scanning that is not present when measuring inclination data alone the rotation of the faceted mirror possibly Nevertheless it has been confirmed that measuring inclination data in both the clockwise and counter clockwise directions noticeably improves the quality Page 48 and reliability of inclination data even for extremely fast scans where the error would be most noticeable 3 4 CASE STUDY O H HINSDALE WAVE RESEARCH LABORATORY The value of inclination sensors use in the field is shown through the following case study In June of 2010 researchers at the O H Hinsdale Wave Research Laboratory HWRL in Corvallis Oregon Figure 6 requested that the Oregon State University geomatics group perform laser scanning for a 3D record of their scale models tested in the tsunami wave basin Figure 6 A view of the scan data collected from three scan positions at the O H Hinsdale Wave Research Lab HWRL tsunami wave basin TWB in Corvallis Oregon In the center of the figure is a scale model of Cannon Beach Oregon located in the southwest corner of the TWB Page 49 3 4 1 Background The HWRL tsunami wave basin is a large fa
80. ols of the surveying profession revolutionizing the industry over the past two decades just as it has impacted our everyday lives Anderson and Mikhail 1998 Recently laser scanning or Light Detection And Ranging LIDAR has emerged as a very reliable efficient and accurate measurement technique with successful implementation in a wide variety of applications including for example CAD model verification e g Son Park and Lee 2002 Ravani et al 2007 historic site and feature preservation e g Bernardini et al 2002 crime scene reconstruction e g Agosto et al 2008 topographic surveying e g Wehr and Lohr 1999 forest resource management e g Wynne 2006 landslide modeling e g Dunning Rosser and Massey 2010 and coastal erosion studies e g Olsen et al 2009 1 1 LASER SCANNING OVERVIEW For a detailed overview of laser scanning the reader is referred to Vosselman and Maas 2010 Only a basic overview is presented herein The modern Page 2 laser scanner is an active sensor that emits a beam of light a laser pulse toward a remote object measures the time of flight to and from the object and computes the distance between the remote object and the scanner While the intricacies of measuring the time of flight vary by system the distance from the scanner to the object the range is calculated using the following equation L 0 5c At 1 1 where L is the distance from the scanner to
81. ontain user manuals for the tools developed to accomplish this research Note that throughout this document the terms coordinate system and reference frame will be used interchangeably Page 2 TRANSFORMATION AND ADJUSTMENT OF POINT DATA The location of a single point in three dimensional space can be described using a Cartesian X Y Z coordinate system Conceptually for example the Rose Garden in Portland Oregon can be described as being a certain distance east X north Y and above Z the Portland State University campus That same point also has another set of Cartesian coordinates measured relative to downtown Portland and yet another set relative to the peak of Mount Hood The point s physical location in space never actually changes but the X Y and Z values of its coordinates will be dramatically different in each coordinate system or reference frame The process of converting point coordinates from one reference frame to another is known as transformation Please note that throughout this section all coordinate systems and rotations will strictly follow the right hand rule see Figure 3 Equations notation and order of operations vary between sources but the theory remains generally the same Least squares equations and transformation equations have been adapted from Ghilani 2010 except where noted Page 8 Figure 3 A right handed coordinate system with positive rotations shown 2 1 TR
82. ore consistent for each setup significantly reducing the overall standard deviation The significant discrepancy in the overview scan data for the leveled setup may appear suspicious but it accurately demonstrates the variability observed in inclination data derived from extremely fast scans as shown in the following section Because of the skew this introduces to the un weighted average an Page 41 average was also computed for each setup from the five slowest scans that was almost identical to the averaged DM results and the TP results suggesting that under the right conditions the inclination sensor is precise and accurate respectively within the standard deviation provided in the specifications From Figure 4 it can be seen that inclination data extracted from a scan with a duration of at least 100 seconds should be reliable for the scanner used in this study This stability threshold will likely vary from scanner to scanner but the same procedure can be used to determine it 3 3 4 Rotation Speed Effects While the stability threshold appears to be correlated with scan duration it remains to be demonstrated that this effect is actually a function of scan speed and not some other factor for example the roll and pitch values such as those presented in the previous section are usually a simple average computed over the duration of the scan and so a smaller sample size in shorter scans may simply be statistically less re
83. own in B and D ccccccceccceeeseeeeeeeeeeeeeeseeeaeeeaeesaees 51 Figure 10 Plot of total station points recorded around the TWB in July 2010 Note that the wavemaker face is on the left in this figure with positive X pointing right and positive Y pointing Up 58 Figure 11 Flowchart detailing the steps toward developing the new HWRL TWB control coordinate system ccccecceeeseeceeeeeeeeeeeeeneees 61 Figure 12 A portion of the wavemaker face colored according to distance from the plane computed in Step 3 Negative orange values indicate that the point was closer to the scan position than HEDIN cose contebau scans tal Seatac sonia Gal ccatena soe tel areca na orc 62 Figure 13 TLS data of the TWB floor from three different scan positions colored according to elevation White points are at Z 2 0 01m Note the concentration of these points close to the scanner that are not observed in the other scan positions 63 Figure 14 East west cross sections CS of the TWB floor separated DY SCAN DOSIMIOMN SP asettce detec E dee acum odes tens 65 Figure 15 North south cross sections CS of the TWB floor separated Dy SC ali POSION Sir a25 ttentenceacaccecssactirecscacieeaica aeeeiee tee a n cients cates 66 Figure 16 Point cloud data set of a model beach at the HWRL after wave inundation Blue represents accretion and orange FEDISSENIS SIOSION rrea EAE 68
84. pared to post registration orientation Inclination sensor data was extracted from overview scans recorded in a 360 window for the specified duration Results indicate a dominant 0 123 rotation around the north south AXIS cccceceeceeceeceeeeeeeeeeeeeeeeneeeeees 52 LIST OF APPENDIX FIGURES Figure Page Figure A1 The control point file loading screen ccccceccceeeeseeeseeeeaeeeaeees 81 Figure A2 Verify that the correct number of points loaded from each file prior t CIIGKING Stal eannan nann AA 81 Figure A3 Adjustment results will be shown on a screen similar to this one Rotation angles are in radians and Degrees Minutes Seconds Copy and paste this output to a blank text file if you want to save it 83 Figure C1 The DriveVZ400 main screen cece cece eeeceeeeaeeeeeeeeeeceeeseeaeeeaes 89 This document is dedicated to my Lord and Savior Jesus Christ who brought me through a tumultuous fall term and without whom would have no reason to leave my home in the first place 1 INTRODUCTION Humans have attempted to measure the earth for millennia a task that continues to grow in complexity and precision to this day Land surveyors are professionals in this field seeking to locate objects in a two or three dimensional reference frame Advancements in computers microprocessors and the ability to measure and define time at very fine resolutions have had a direct impact on the to
85. placed in locations where it was safe to get to have also not used the compensator when mapping cliff faces to 9 3 10 3 15AM maximise the available tidal window this was also as sufficient wide ranging control stations and targets could be established at the extremeties of the scan to allow for reliable verticality controls
86. r blunders were made and systematic error has been minimized or eliminated any effort to match the two data sets together optimally would evenly distribute the error among the input parameters By definition least squares adjustment methods produce the most mathematically correct match between two data sets by minimizing the sum of the square of the errors Ghilani 2010 A simple average is a special case of least squares that can be applied to measurements of a single quantity The least squares approach requires equations to be linearized and requires iterative solutions when applied to nonlinear equations such as the transformations discussed previously Han 2010b developed a noniterative alternative to least squares methods using eigentheory known as the noniterative solution to linear transformations or NISLT Han Guo and Chou 2011 Page 15 Fundamentally both methods attempt to compute the seven parameters for a three dimensional similarity transformation introduced in Section 2 1 that would result in the closest match of one data set to another The following sections will introduce the basic concepts of both methods and compare their strengths and weaknesses See also Ghilani 2010 Han 2010a Han 2010b and Han Guo and Chou 2011 2 2 1 Least Squares Adjustment General Case Given a set of linear equations relating an equal number of independent unknowns X4 X2 Xn a unique solution can
87. ram immediately requests the user to select a CSV or TXT file with control data to remain fixed and then again for measured data to be adjusted These files must have the same number of points Jiu My Computer J sivia ld arcgis Ji ie5tmp Riegl Scans Type Date Modified File Folder 7 282 a af PM File Folder 5 13 2 59 PM File Folder 10 20 u38 PM Page 81 dichato_ctrl tet dichato_meas tet logfile tet WaveFaceCombined tt 626 bytes tt File 5 16 2 33 AM 589 bytes tt File 5 16 2 57 AM Obytes tt File 5 19 2 00 PM 44 6 MB tet File 5 19 2 18 PM File name dichato_ctrl txt res of pe Figure A1 The control point file loading screen If loading succeeded then the user may start the adjustment by clicking Start Importing 19 points from J dichato_ctrl txt 19 points successfully imported Importing 19 points from C temp dichato_meas txt 19 points successfully imported Figure A2 Verify that the correct number of points loaded from each file prior to clicking Start Page 82 The program currently seeds the iterative process with rotations and translations of 0 and a scale factor of 1 This may be an issue for some data sets that will be unable to converge without good initial values The program will iterate up to 200 times or until the new RMS has changed less than 0 000001 from the previous iteration whichever comes first The rotation angles are forced to be between
88. rection and ten were recorded in a counter clockwise direction The raw inclination data from each scan was software extracted and averaged Inclination data was also measured from the same setup using the semi automatic Riegl VZ 400 inclination measurement methods described in Table 3 3 3 6 2 Scanner directional effects assessment results and discussion The results of the three sets of tests are summarized in Table 5 Scans all represents the combination of 20 scans Scans paired was generated by averaging the 10 pairs of consecutive clockwise and counter clockwise scans Page 46 Scans cw only represents only the 10 clockwise scans while Scans ccw only represents only the 10 counter clockwise scans Discrete represents a set of 20 measurements using method 2 from Table 3 Continuous represents a set of 20 measurements using method 3 from Table 3 Table 5 Comparison of inclination measurement methods in degrees cw clockwise and cew counter clockwise Method Average stDev Range Scans Roll 0 753 0 103 0 379 all Pitch 0 826 0 043 Scans Roll 0 753 0 049 paired Pitch 0 826 0 013 0 036 Scans Roll 0 823 0 073 0 246 cwonly Pitch 0 790 0 014 0 042 Scans Roll 0 683 0 078 0 188 ccwonly Pitch 0 862 0 026 0 075 0 789 0 007 0 027 Pitch 0 823 0 008 Roll 0 787 0 005 Conti aac ontinuous Steh T ogs 0 001
89. s Page 30 3 2 2 Inclination Sensors Most modern laser scanners now have inclination sensors JADA Productions 2010 also referred to as level compensators inclinometers or tilt Sensors For clarity we will refer to them herein solely as inclination sensors Given a right handed scanner coordinate system with the X axis pointing along the scanner s initial line of sight the Y axis at 90 counter clockwise from the X axis in the horizontal plane and the Z axis pointing vertically upwards rotation of the scan setup into a project coordinate system X Y Z is usually described in terms of roll pitch 6 and yaw y as rotations around the x Y and Z axes respectively Note that these three rotations can also be summarized as a single rotation about a single vector quaternion Inclination sensors measure roll and pitch of the instrument while yaw is related to the scanner s bearing or azimuth relative to an external project coordinate system Given a laser scan data point Po with coordinates X Y Z alignment adjustments can be applied by rotating the point around the X Y and Z axes centered at the scan origin P R R R Po 3 1 P R a B y Po 3 2 Page 31 R a f y cosp cosy cosasiny sinasinfbcosy sinq siny cosasinf cosy cos f siny cosacosy sinasinfsiny singa cosy cos sinf siny 3 3 sin p sin amp cos fp COS acos p Where Rz Ry and Rz are the rotation matrices aroun
90. scanner head rotation positions polar axis Scan durations listed are approximate Measurements were taken approximately ONCE per Second ccc ccc ceccecceeceeceeceeceeceeceeeeeeeeeseeseeseeseeseeseeseeseeees Figure 6 A view of the scan data collected from three scan positions at the O H Hinsdale Wave Research Lab HWRL tsunami wave basin TWB in Corvallis Oregon In the center of the figure is a scale model of Cannon Beach Oregon located in the southwest CONE ONG VV Be isnan a essere Aeetacntan Figure 7 A typical HWRL target colored by return signal amplitude left and a grayscale photograph of the same target right Figure 8 The HWRL tsunami wave basin building walls are outlined the wave basin itself is shaded Scan positions are denoted SP with an arrow indicating setup orientation Brass floor benchmarks are denoted BM ccc cccecceccecceccccececcecceauecceaneas Page side 3 sites 4 aie 8 an 40 re 43 ui 48 hes 49 ends 50 LIST OF FIGURES Continued Figure 9 Representative iso views and cross sections 40x vertical exaggeration of the wave lab roof after surface registration of scan position 2 light red to scan position 7 dark black Scan position 2 is approximately 7cm below scan position 7 on the west end and 7cm above it on the east end as shown in A and C Once the control had been corrected the scans aligned properly as sh
91. slation vector and the rotation angles 1 i fz are the same as defined in Han 2010 for the similarity model However it was observed that R is not a true rotation matrix and as such the rotation matrix computed using the rotation angles R calc will not be the same That calculated rotation matrix the new translation vector derived from it and the new set of transformed coordinates are also provided in addition to the typical results from the NISLT technique A Root Mean Square Error is also calculated using the results of both techniques Page 8 APPENDIX C SCANNER INTERFACE PROGRAM NOTES Page 88 DriveVZ400 User Manual 5 19 2011 By Evon Silvia Graduate Research Assistant School of Civil and Construction Engineering evon silvia qmail com Introduction The Drive VZ400 tool is a convenient lightweight scan collection interface that can run by itself or within other programs It cannot process data it can only record It can be used on any windows based device with networking capabilities and all of the information needed to record a scan is conveniently displayed on one screen It also Supports automatic or command line operation several other viable options exist to operate the RIEGL VZ 400 scanner e On board interface This is the most straightforward manner but the menu structure can be counter intuitive and inconvenient Simple operations like entering a project name require pressing buttons repeatedly
92. te the concentration of these points close to the scanner that are not observed in the other scan positions Page 64 The white patch surrounding a scan position does not appear in the corresponding area in the other two scans This suggests that points close to the scanner have a slight approximately 5 mm bias that makes them appear closer than they otherwise would As the scanner used in this study was calibrated as a long range gt 100 meters scanner this explanation for the 5 4 millimeter discrepancy in Step 6 of the previous section seems plausible This becomes especially clear while comparing the cross sections in Figure 14 east west and Figure 15 north south to the relative scan position locations shown in Figure 13 Prior to the 5 4 millimeter correction in Step 6 the total station data orange circles were consistently below the scan data even after accounting for the noticeable variations in the TWB floor itself The numerical least squares adjustment corrected this error to shift the control up such that the total station data would on average align with the scan data Using either a coordinate system such as the one developed in this section or by implementing the leveling methods detailed in the Chapter 3 laser scan registration can be done rapidly and easily with minimal user input Page 65 dS uoilisod ueds Aq pajyeiedas 100 ML y JO SD SUOI D9BS SSOID JS9M SeJ yp ainbi4 MOIEIS ERO 7S0 E
93. the object Ca is the speed of light under the current atmospheric conditions 299 792 458 m s in a vacuum and At is the time of flight of the laser pulse Scanner measurement frequencies are typically on the order of tens of thousands of pulses per second JADA Productions 2010 conducted while simultaneously rotating the scanner mirror along a given range of horizontal and vertical y angles Figure 1 Some newer scanners can conduct full waveform analysis enabling the recording of multiple returns per pulse The set of 3D points collected from one or more scans Is referred to as a point cloud Page 3 Scanner Figure 1 Scanners rotate horizontally o and vertically y recording thousands of points per second All laser scanners are also line of sight instruments meaning that they can only measure the visible surfaces of objects as seen from the setup location Therefore multiple scans are usually required for full coverage of an area or object Figure 2 demonstrates the use of two additional scan setups to fill gaps black regions present in the original scan left due to occlusions Because each setup or scan position has its own spherical coordinate system based on the location and orientation of the scanner each scan must Page 4 be shifted and rotated to be aligned to a common reference frame This process is known generally as scan registration or georeferencing if a geographic coordinate system is
94. them being time stamped with the date and time in GMT of the scan as its name The files are stored ina riproject folder with the projects name in front and in the corresponding scanPos folder for the current scan position Note that the RIPROJECT folder will Page 92 be either stored on the scanner s internal hard drive or on the USB drive plugged into it depending on your settings These are the files usually stored in the ScanPos folder upon scan completion YYMMDD_HHMMSS rxp the binary raw scan data itself with housekeeping and monitor data included YYMMDD_HHMMSS mon rxp a small sample of the raw scan data Useful for real time change analysis and verifying that the correct area was scanned YYMMDD_HHMMSS png a small image file that displays the scan in a polar grid Useful for easy visual inspection of the scan without having to load the raw data YYMMDD_HHMMSS pat an ASCII file with the scan parameters YYMMDD_HHMMSS prv a preview file that believe only RISCAN can use YYMMDD_HHMMSS_Image jpg scan photos taken along the horizontal range of the scan Some parameters can be read by an ASCII text reader Command line operation The DriveVZ400 program can be run from the command line for automated operation with or without user input This is accomplished by a series of arguments separated by spaces following the executable in a form similar to the following C programs DriveV
95. uly 13 2010 high resolution 360 scans of the empty and mostly dry TWB were collected from three leveled scan positions above the stability threshold Five targets Figure 7 distributed around the TWB on supporting columns were scanned at fine resolution from each scan position Coordinates obtained from a total station were collected in an arbitrary coordinate system Page 59 for every target and benchmark across the wavemaker face and in four cross sections on the floor Figure 10 Following collection in the field the point cloud from each scan setup was leveled programmatically using its inclination values Total station coordinate data was processed and treated as control data due to its very high precision Digital leveling data was also collected where possible to verify elevations collected with the total station In all cases the total station and digital level elevation values agreed within 1 millimeter Finally target tiepoints were extracted from the fine resolution scans for each setup in preparation for tiepoint registration 4 1 2 Development of the New TWB Control Coordinate System The six step development of the new HWRL TWB control coordinate system is detailed in Figure 11 Between each step the scans were aligned using tiepoint registration a least squares adjustment technique that matches the scan based tiepoints to the updated total station control points In summary the key steps are 1 Shift to
96. vector components in the rotated reference frame along the axes of the original reference frame Property 3 the usefulness of which will be shown in Section 2 2 5 can be written explicitly as ae ho fae 1 2 2 1 1o2 13 1 2 2 i 134 132 133 1 Page 12 T 121 7317 1 2 2 Oe ee ioe fee 1 2 6 After rotation Pr is then scaled and translated relative to the rotated reference frame to complete the transformation 2 1 2 Scale scaling point coordinates can follow either a similarity model with a uniform scale factor or an affine model where scale factors for each axis must be determined scale factors in TLS are usually used to model atmospheric effects and can also be used to convert topographic measurements to grid distances with a grid scale factor Atmospheric corrections to the range values for the instrument used in this study are modeled using the following guidelines and standard values RIEGL LMS GmbH 2011 e A change in temperature of 1 C yields a change in correction of 1 ppm calibrated at 12 C e A change in air pressure of 10 millibars yields a change in correction of 2 7 ppm calibrated at 1000 mbar e A change in relative humidity from 0 to 100 yields a change in correction of only 0 5 ppm calibrated at 60 Page 13 While the instrument used in this study automatically applies atmospheric corrections during measurement this may not alw
97. verview at 0 03 increments o default extremely fast 360 overview good for testing code e p P alternate scan pattern This is the scan window parameters that set up the actual scan range separated by commas vstart vstop vincr hstart hstop hincr o vstart vertical start position 30 to 130 0 001 precision o vstop vertical stop position same o vincr vertical angle increment 0 001 to 0 098750 0 000001 precision theoretically 0 0001 practical o hstart horizontal start position 0 to 360 0 001 precision o hstop horizontal stop position 0 to 720 0 001 precision o hincr horizontal increment 0 001 to 0 499 same precision as vertical e i l images requested true false toggle of whether or not images should be recorded o 0 do not record images after the scan o 1 if possible record images after the scan o 2 under develooment do HDR photography e r R operate in real time mode You will always have this marked 1 o OQ operate normally recording data as usual o 1 record roll pitch values prior to the scan download the monitor file and write an output file to the executable s directory Real time mode If the user elects to run in real time mode an ASCII file named vz_info txt will be created in the same directory as the executable This file can be parsed by another program upon scan completion to retrieve information needed for real time analysis The tags and information
98. ward developing the new HWRL TWB control coordinate system Page 62 4 1 3 Additional Considerations The wavemaker face shows significant variability 0 05 meters in the X direction due to its construction as a series of independently controlled panels and warping of the panels themselves Figure 12 While this variation went essentially undetected in the ten total station data points it was clear upon examination of the LIDAR data demonstrating the value of TLS for rapidly recording complex geometric shapes However the high precision total station data did provide a reliable check that enabled the correction of the 1 35 centimeter error observed in Step 5 The source of this error appears to be saturation effects due to scan position 1 being too close to the wavemaker face a highly reflective surface Vosselman and Maas 2010 5 0cm 2 5cm 0 0cm 2 5cm 5 0cm Figure 12 A portion of the wavemaker face colored according to distance from the plane computed in Step 3 Negative orange values indicate that the point was closer to the scan position than the plane The 5 millimeter discrepancy between the TLS data and total station data observed in Step 6 led to an investigation to uncover its cause The TWB floor as scanned from three different positions is shown in Figure 13 Page 63 Figure 13 TLS data of the TWB floor from three different scan positions colored according to elevation White points are at Z 2 0 01m No
99. while the program is running Note that the conversation with the scanner is stored in C Temp logfile txt for debugging purposes and can be opened after scan completion with Open Logfile At any time you can pop out the Info Log by double clicking its title bar Return it by double clicking again The interface is organized in a logical flow project setup site conditions scan settings and options Only the scan settings need change between scans for most sites Project setup Project name All projects are stored on the scanner or a USB drive plugged into it Type in a new project name in the entry box or download a list of existing projects from the scanner and select one Some notes e All projects have a riproject extension The program will not recognize projects without that extension e Existing projects will have a tag appended to their name indicating where the project was detected These tags are not part of the project name o internal indicates that the project was found on the scanner s internal drive o usb indicates that the project was found on the USB drive e The Debug option is meant for offline troubleshooting purposes and should not be used in normal operation Storage Media Select where you want the project and scan data to be stored USB Internal is recommended for nearly all projects it attempts to save on the USB but will save on the internal drive if no USB driv
100. with scanner accuracy A comparison of Set 1 roll and pitch values before and Page 52 after registration is shown in Table 6 Results indicated a consistent rotation around the north south axis of approximately 0 12 degrees from level Table 6 Comparison of inclination data from inclination sensors compared to post registration orientation Inclination sensor data was extracted from overview scans recorded in a 360 window for the specified duration Results indicate a dominant 0 12 rotation around the north south axis bi Inclination Tiepoints Roll Pitch Duration at Roll Pitch SD m Roll mA Yaw L N S Average SD To ensure that the scan inclination readings were correct and that the problem was inherent to the control the team returned to the site with a survey grade total station to verify the relative positions of the benchmarks and TLS targets as well as to provide cross sections of the empty basin floor A digital automatic level was used to confirm elevation readings made with the total station to ensure that the basin floor and scan data were level 3 4 4 Results Further investigation revealed that the Southeast Benchmark elevation on record was approximately 75mm higher than observed with a survey grade level which alone accounts for approximately 0 11 degrees of the observed error Attempts to explain the cause of such a large error would be Page 53 specul

View/Open - Oregon State University

Contents

Download Pdf Manuals

Related Search

Related Contents