1 / 69

The City Scanning Project :

The City Scanning Project :. Validation and parallel algorithms. Olivier Koch, Visitor from March to August 2002. Advisor : Seth Teller. Contents. 1/ Project overview 2/ code validation 3/ parallel algorithms 4/ a few ideas.

romney
Télécharger la présentation

The City Scanning Project :

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. The City Scanning Project : Validation and parallel algorithms Olivier Koch, Visitor from March to August 2002 Advisor : Seth Teller

  2. Contents 1/ Project overview 2/ code validation 3/ parallel algorithms 4/ a few ideas The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  3. Project overview Reconstruct a realistic 3D model of an urban environment from calibrated real-world images Technology Square Argus The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  4. Project overview • Nodes The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  5. Project overview • Edge detection The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  6. Project overview • assume edges are horizontal • determine corresponding orientation in 3D world edge Camera plane Camera center Ground plane The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  7. Project overview • assume edges are horizontal • determine corresponding orientation in 3D world edge Camera plane Camera center Ground plane The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  8. Project overview • assume edges are horizontal • determine corresponding orientation in 3D world edge Camera plane Camera center Ground plane The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  9. Project overview • assume edges are horizontal • determine corresponding orientation in 3D world Horizontal plane Camera plane Camera center Ground plane The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  10. Project overview • assume edges are horizontal • determine corresponding orientation in 3D world Horizontal line Horizontal plane Camera plane Camera center Ground plane The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  11. Project overview • assume edges are horizontal • determine corresponding orientation in 3D world Horizontal line Horizontal plane Camera plane Camera center Ground plane The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  12. Project overview • assume edges are horizontal • determine corresponding orientation in 3D world horizontal line horizontal line Horizontal plane Camera plane Camera center Ground plane The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  13. Project overview • assume edges are horizontal • determine corresponding orientation in 3D world Horizontal line Horizontal plane Camera plane Camera center Ground plane The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  14. Project overview • assume edges are horizontal • determine corresponding orientation in 3D world Horizontal line Camera center Ground plane The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  15. Project overview • assume edges are horizontal • determine corresponding orientation in 3D world Horizontal line Ground plane The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  16. Project overview • assume edges are horizontal • determine corresponding orientation in 3D world Z Horizontal line X  Y Ground plane  = azimuth of 3D edge The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  17. Project overview • Edge detected… The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  18. Project overview • Edge detected… now have an azimuth ! • each azimuth is color-coded. The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  19. Project overview • histogram edges weight azimuth 0 360 Façade orientation wrt due east The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  20. Project overview • histogram horizontal edges weight azimuth 0 360 The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  21. Project overview • histogram horizontal edges weight azimuth 0 360 The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  22. Project overview • keep orientations that are supported by at least 3 nodes weight azimuth  0 360 The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  23. Project overview • determine corresponding edges weight azimuth  0 360 The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  24. Project overview From : Extracting Textured Vertical Facades from Controlled Close-Range Imagery, S. Teller, S. Coorg in Proc. CVPR, pp. 625-632, June 1999. The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  25. Project overview correlation computed correlation value step grid cell projected edge The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  26. Project overview correlation peak step grid cell sweep step The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  27. Project overview correlation peak step grid cell sweep step The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  28. Project overview correlation peak step grid cell The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  29. Project overview The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  30. Project overview The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  31. Project overview The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  32. Project overview The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  33. Project overview 2 2 2 1 1 1 1 1 1 The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  34. Project overview The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  35. Project overview The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  36. Project overview The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  37. Project overview The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  38. Contents 1/ Project overview 2/ code validation 3/ parallel algorithms 4/ a few ideas The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  39. Code validation • edge detection improvement automatic edges manual edges The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  40. Code validation • edge detection improvement manual edges automatic edges The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  41. Code validation • polygon filtering : remove polygons without support. without filter with filter The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  42. Code validation • debugging bugs warnings CVD Rational Purify log files, class constructors, … The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  43. Contents 1/ Project overview 2/ code validation 3/ parallel algorithms 4/ a few ideas The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  44. Parallel algorithms not processed • why parallelize ? being processed processed IRIX machine grid cells The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  45. Parallel algorithms not processed • why parallelize ? being processed processed IRIX machine The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  46. Parallel algorithms not processed • why parallelize ? being processed processed IRIX machine The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  47. Parallel algorithms not processed • why parallelize ? being processed processed > 7 hours on IRIX single processor = The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  48. Parallel algorithms • first method Machine 2 Machine 1 Machine 3 Machine 4 Machine 5 The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  49. Parallel algorithms • first method • CPU speed varies across machines • time to process a cell is not constant difficult optimization problem… Static allocation turpentine ray trace mosaic acetone hue panorama orange 2285 sec. 2678 2443 2131 2711 2342 1755 1939 Computing time : 45 min. Standard dev : 3.4 % The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

  50. Parallel algorithms • second method Machine 2 Machine 1 Machine 3 Machine 4 Machine 5 The City Scanning Project – validation and parallel algorithms – July 2002 – koch@graphics.lcs.mit.edu

More Related