Acta Optica Sinica, Volume. 39, Issue 10, 1015002(2019)

Research and Implementation of Industrial Photogrammetry Without Coded Points

Jun Yan, Nan Ye*, Tingcheng Li, and Hongyu Zhu
Author Affiliations
  • College of Mechanical and Electrical Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing, Jiangsu 210016, China
  • show less
    Figures & Tables(23)
    Schematic of photogrammetry without coded points
    Process of multi-view structure reconstruction without coded points
    White basecoat of digital speckle pattern
    Pseudocode of random growth method of speckle particle
    Speckle particles with random shapes
    Speckle patterns with different sizes of speckle particles. (a) 5-8 pixel; (b) 9-12 pixel; (c) 13-20 pixel
    Speckle patterns with different proportions of speckle particles. (a) 20%; (b) 40%; (c) 60%
    Speckle patterns with different grayscale ranges (0-255) of speckle particles. (a) 40-100; (b) 20-80; (c) 0-60
    DIC matching results of projected speckles. (a) Left image; (b) right image
    Mismatching problem of DIC between images with large rotation angle. (a) Left image; (b) right image
    Matching strategy of RFDIC
    Matching results of RFDIC and Patch Matching methods. (a) Reference image; (b) RFDIC method; (c) Patch Matching method
    Pseudocode of multi-view geometry reconstruction without coded points
    Sheet metal of aircraft skin
    Experimental scene of photogrammetry without coded points
    Multi-view structure reconstructed by photogrammetry without coded points
    Point cloud and triangular mesh model reconstructed from two views
    Comparison of overall deviation of point cloud
    • Table 1. Comparison of matching results between RFDIC and Patch Matching methods

      View table

      Table 1. Comparison of matching results between RFDIC and Patch Matching methods

      AlgorithmNumber of matching pointMatching rate /%Correlation coefficient
      RFDIC3142100>0.99
      Patch Matching3142100>0.99
    • Table 2. Comparison of running time and point cloud error between RFDIC and Patch Matching methods

      View table

      Table 2. Comparison of running time and point cloud error between RFDIC and Patch Matching methods

      Running time/sAverage error /mmStandard deviation /mm
      RFDICPatch Matching
      1.6771.150.05480.0096
    • Table 3. Reprojection mean error and RMSE of multi-view structure

      View table

      Table 3. Reprojection mean error and RMSE of multi-view structure

      MethodAverage error /pixelRMSE /pixel
      SIFT0.1240.205
      RFDIC0.0530.074
    • Table 4. Reconstruction results and error comparison of scale bars’ length

      View table

      Table 4. Reconstruction results and error comparison of scale bars’ length

      MethodScale bar to recover scaleScale bar to be verifiedStandard distance /mmReconstructed distance /mmError /mm
      SIFT12860.916860.799-0.117
      21861.115861.2310.116
      Code points12860.916860.912-0.004
      21861.115861.1180.003
      RFDIC12860.916860.9210.005
      21861.115861.110-0.005
    • Table 5. Comparison of point cloud deviations reconstructed from two views

      View table

      Table 5. Comparison of point cloud deviations reconstructed from two views

      DirectionMaximum /mmAverage /mmStandard deviation /mm
      Negative-0.4605-0.05590.0475
      Absolute0.46050.05420.0464
      Positive0.43790.05210.0450
    Tools

    Get Citation

    Copy Citation Text

    Jun Yan, Nan Ye, Tingcheng Li, Hongyu Zhu. Research and Implementation of Industrial Photogrammetry Without Coded Points[J]. Acta Optica Sinica, 2019, 39(10): 1015002

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Machine Vision

    Received: Apr. 10, 2019

    Accepted: Jun. 21, 2019

    Published Online: Oct. 9, 2019

    The Author Email: Ye Nan (yen@nuaa.edu.cn)

    DOI:10.3788/AOS201939.1015002

    Topics