IGARSS, 24-29 July 2011, Vancouver, Canada (Session FR2.T03) <br />Quality Assessment for LIDAR Point Cloud Registration u...
NTUCvE Surveying Engineering Group<br />Outline<br /><ul><li>Introduction
Using In-Situ Conjugate Features
Weighted NISLT Approach
Quality Assessment
Numerical Validation
Conclusion</li></li></ul><li>Introduction<br /><ul><li>Light Detection and Ranging (LIDAR) is capable of acquiring 3D spat...
 Can be equipped on platforms of various kinds (air-borne, mobile, and terrestrial).
Usually requires multiple scans in order to construct a complete and accurate 3D model.</li></ul>Reason 1: Incompleteness ...
Introduction (cont’d)<br /><ul><li>Incompleteness due to obstructions</li></ul>Many obstructions could occur when the LIDA...
Introduction (cont’d)<br /><ul><li>Error magnification due to projective geometry</li></ul>Point coordinates are based on ...
Introduction (cont’d)<br /><ul><li>Registration of LIDAR datasets from multiple stations </li></ul>Each dataset is defined...
Using In-Situ Features<br />Obtaining the transformation parameters<br /><ul><li>Classic approach: point-based least-squar...
 Find (>=3) conjugate points in two LIDAR datasets
 Perform least-squares parameter estimations</li></ul>Requires extra effort to set up identifiable targets (e.g. control s...
Using In-Situ Features<br />Obtaining the transformation parameters<br /><ul><li>Proposed approach: using directly in-situ...
Using In-Situ Features<br /><ul><li>In-situ features usable for LIDAR dataset registrations</li></ul>Highway surfaces     ...
Weighted NISLT Approach<br /><ul><li>Once feature correspondence is established, the transformation parameters are estimat...
Weighted NISLT Approach<br />Rotational parameters<br />where ΔX and ΔX’ are the matrices by stacking all the normalized r...
Weighted NISLT Approach<br /><ul><li>Characteristics of weighted NISLT approach</li></ul>     - Closed-form solution, requ...
Quality Assessment<br /><ul><li>Classical point-based approach: </li></ul>Registration quality is typically evaluated by t...
Quality Assessment<br /><ul><li>Proposed approach: </li></ul>Here features of various kinds are used for a registration. T...
(a)		       (b)<br />(c)	    	      (d)<br />Quality Assessment<br /><ul><li>Interpretation of a registration solution: </...
S2<br />S1<br />Numerical Validation<br /><ul><li>Data collection: </li></ul>A case study was performed for a 250m-long re...
Upcoming SlideShare
Loading in …5
×

QUALITY ASSESSMENT FOR LIDAR POINT CLOUD REGISTRATION USING IN-SITU CONJUGATE FEATURES

664 views

Published on

Published in: Technology, Business
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
664
On SlideShare
0
From Embeds
0
Number of Embeds
16
Actions
Shares
0
Downloads
10
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

QUALITY ASSESSMENT FOR LIDAR POINT CLOUD REGISTRATION USING IN-SITU CONJUGATE FEATURES

  1. 1. IGARSS, 24-29 July 2011, Vancouver, Canada (Session FR2.T03) <br />Quality Assessment for LIDAR Point Cloud Registration using In-Situ Conjugate Features<br />Jen-Yu Han1, Hui-Ping Tserng1, Chih-Ting Lin2<br />1 Department of Civil Engineering, National Taiwan University<br />2 Graduate Institute of Electronics Engineering, National Taiwan University<br />
  2. 2. NTUCvE Surveying Engineering Group<br />Outline<br /><ul><li>Introduction
  3. 3. Using In-Situ Conjugate Features
  4. 4. Weighted NISLT Approach
  5. 5. Quality Assessment
  6. 6. Numerical Validation
  7. 7. Conclusion</li></li></ul><li>Introduction<br /><ul><li>Light Detection and Ranging (LIDAR) is capable of acquiring 3D spatial information in a fast and automatic manner.
  8. 8. Can be equipped on platforms of various kinds (air-borne, mobile, and terrestrial).
  9. 9. Usually requires multiple scans in order to construct a complete and accurate 3D model.</li></ul>Reason 1: Incompleteness due <br /> to obstructions<br />Reason 2: Error magnification due <br /> to projective geometry<br />
  10. 10. Introduction (cont’d)<br /><ul><li>Incompleteness due to obstructions</li></ul>Many obstructions could occur when the LIDAR point cloud is collected from a single station.<br />Only partial information is acquired for the 3D object.<br />
  11. 11. Introduction (cont’d)<br /><ul><li>Error magnification due to projective geometry</li></ul>Point coordinates are based on range and angular measurements both of which contain errors.<br />As a result, the quality will become lower for outer regions. <br />
  12. 12. Introduction (cont’d)<br /><ul><li>Registration of LIDAR datasets from multiple stations </li></ul>Each dataset is defined in an arbitrary local reference frame.<br />A 3D similarity transformation model is usually postulated to relate the datasets defined in different reference frames. <br />1<br />2<br />2<br />1<br />2<br />s: scale<br />R: rotation matrix<br />t: translation vector<br />1<br />Station 1 Station 2<br />
  13. 13. Using In-Situ Features<br />Obtaining the transformation parameters<br /><ul><li>Classic approach: point-based least-squares approach
  14. 14. Find (>=3) conjugate points in two LIDAR datasets
  15. 15. Perform least-squares parameter estimations</li></ul>Requires extra effort to set up identifiable targets (e.g. control spheres or reflective sticks) or perform feature extractions.<br />Requires a set of good initial values and iterative computations to obtain reliable parameter estimates.<br />
  16. 16. Using In-Situ Features<br />Obtaining the transformation parameters<br /><ul><li>Proposed approach: using directly in-situ features</li></ul> Extended feature types<br /><ul><li> Definite features</li></ul> Points: vectors between points<br /> Lines: directional vectors <br /> Planar patches: normal vectors<br /><ul><li>Indefinite features</li></ul> Groups of points: eigenvectors of <br /> the tensor field constructed by a <br /> group of point.<br />With these extended feature types, it becomes possible to use the geometric components that are already inherent in the scanned object. <br />
  17. 17. Using In-Situ Features<br /><ul><li>In-situ features usable for LIDAR dataset registrations</li></ul>Highway surfaces Bridge pillars<br />Slope surfaces and edges Structure edges and rails<br />No need to set up control targets  reduce the cost for field work. <br />
  18. 18. Weighted NISLT Approach<br /><ul><li>Once feature correspondence is established, the transformation parameters are estimated by the weighted NISLT (Non-Iterative Solutions for Linear Transformations) technique:</li></ul>Scale parameter<br />where dxij and dx’ij are coordinate differences (vectors) in the original and transformed systems, is the weight matrix, lkis a kx1 unity vector. <br />
  19. 19. Weighted NISLT Approach<br />Rotational parameters<br />where ΔX and ΔX’ are the matrices by stacking all the normalized row vectors in the original and transformed systems. <br />Translational parameters<br />
  20. 20. Weighted NISLT Approach<br /><ul><li>Characteristics of weighted NISLT approach</li></ul> - Closed-form solution, requires no initial<br />values nor iterative computations  <br />highly efficient compared to <br />LSQ-based approaches.<br /> - Weighted parameter estimation model  uncertainties of input <br /> observables can be realistically taken into consideration.<br /> - Accepts input observables of different kinds (e.g. vectors between <br /> points, directional vectors of linear features, normal vectors of <br /> planar features, and eigenvectors of groups of points)  make <br /> possible a direct use of various in-situ geometric features. <br />
  21. 21. Quality Assessment<br /><ul><li>Classical point-based approach: </li></ul>Registration quality is typically evaluated by the post-fit residuals for point coordinates after applying the estimated parameters.<br /> : post-fit residual vector of point i<br />n : number of conjugate points<br />This index gives a vague interpretation on the obtained result since it represents only the positional agreement between two datasets  geometrical similarity is not considered!!<br />
  22. 22. Quality Assessment<br /><ul><li>Proposed approach: </li></ul>Here features of various kinds are used for a registration. The quality is then evaluated based on the following two indexes:<br />Absolute Consistency (qa) Relative Similarity (qr)<br />Positional alignment Geometric similarity<br />: post-fit residual vector of conjugate point i or the vector between point i ‘s <br /> projected points on two conjugate features. <br />: the angle between two conjugate vectors (directional vectors, normal <br /> vectors, or eigenvectors) after the registration. <br />: the numbers of conjugate points and conjugate vectors<br />
  23. 23. (a) (b)<br />(c) (d)<br />Quality Assessment<br /><ul><li>Interpretation of a registration solution: </li></ul>Moderate qa, good qr. <br />Moderate qa and qr. <br />Poor qa, good qr. <br />Poor qa and qr.<br />The quality of a registration solution can be explicitly defined by the proposed two indexes qa and qr.<br />
  24. 24. S2<br />S1<br />Numerical Validation<br /><ul><li>Data collection: </li></ul>A case study was performed for a 250m-long reinforced concrete (RC) bridge in Taipei City.<br />Two LIDAR stations (S1, S2) were set up about 80m away from the bridge.<br />
  25. 25. Numerical Validation<br /><ul><li>Data collection (cont’d): </li></ul>LIDAR point cloud was collected at each station using a Trimble® GS200 Terrestrial Laser Scanner.<br />Resolution for the scanned points of the bridge was roughly between 0.02m ~ 0.04m.<br />No control sphere or reflective stick was set up in the scanned area. <br />TrimbleGS200 Laser Scanner<br /> - Range: 2m~200m<br /> - Accuracy: range = 6 mm @ 100 m<br />angular = 6 mm @ 100 m<br /> - Max. Density: 3mm@100m<br />
  26. 26. Numerical Validation<br /><ul><li>Collected datasets and in-situ features used for registration</li></ul>Two sets of LIDAR point clouds were collected at the two stations.<br />Since no control point was available, in-situ features were selected from the datasets and used for a registration.<br />Two pillars, a rail and a beam surface were used as conjugate features. <br />Station 1 Station 2<br />
  27. 27. Numerical Validation<br /><ul><li>NISLT registration</li></ul>The eigenvectors of conjugate features were used as observables while solving for the transformation parameters based on the proposed weighted NISLT approach.<br />Station 1 Station 2<br />
  28. 28. Numerical Validation<br /><ul><li>Registration results (integrated point clouds)</li></ul>Shown in true colors<br />Shown in blue for points collected at station 1 and in red for points collected at station 2<br />
  29. 29. Numerical Validation<br /><ul><li>Registration results (integrated point clouds)</li></ul>S2<br />S1<br />Integrated<br />
  30. 30. Numerical Validation<br /><ul><li>Registration results (quality assessment)
  31. 31. Absolute consistency (qa) = 3.81cm.
  32. 32. Relative similarity (qr) = 1.864e-4 .
  33. 33. qr is equivalent to a 3.73cm positional distortion for an object of </li></ul> size 200m. Equally accurate in terms of positional agreement and <br /> geometric similarity.<br /><ul><li> Both values are within a reasonable range considering the </li></ul>2cm~4cm resolution of the original LIDAR datasets  the <br /> registration quality is mostly dependent on the point resolution in <br /> this case.<br />
  34. 34. Conclusion<br /><ul><li>The proposed approach increases the number of usable features for a registration solution  the cost for LIDAR field work can be significantly reduced.
  35. 35. The weighted NISLT enables an efficient parameter estimation when in-situ hybrid conjugate features are used.
  36. 36. The two quality indexes (absolute consistency and relative similarity) give a complete and explicit quality indication for a registration solution.
  37. 37. An automatic approach for selecting qualified in-situ features should be developed in the future.</li></li></ul><li>Thanks for your attention<br /> For more information, please contact:<br /> Jen-Yu Han, Ph.D.<br /> Department of Civil Engineering, National Taiwan University<br /> Email: jyhan@ntu.edu.tw Phone: +886-2-33664347<br /> Website: http://homepage.ntu.edu.tw/~jenyuhan<br />

×