Journals Articles Conferences News About CLP
Submit a paper Email Alert
Search
Search
Articles
Journals
News
Advanced Search
Top Searches
2D MaterialsTransformation opticsQuantum PhotonicsSmall lasersSemiconductor UV PhotonicsSupersymmetric microring laser arrays

Optics and Precision Engineering, Volume. 29, Issue 9, 2255(2021)

Sparse mixture iterative closest point registration

Yue-sheng LIU*, Xin-du CHEN, Lei WU, Yun-bao HUANG, and Hai yan LI
Author Affiliations
  • College of Electromechanical Engineering， Guang Dong University of Technology.， Guangzhou510006， Guangdong
    • show less
    AbstractGet PDF(in Chinese)
    Figures&Tables (19)
    Equations (22)
    References (21)
    Tools
    Paper Information
    Topics
    Figures & Tables(19)
    Diagram of partial overlap point-sets registration

    Fig. 1. Diagram of partial overlap point-sets registration

    Download full size
    View in Article
    Sigmoid function with different median

    Fig. 2. Sigmoid function with different median

    Download full size
    View in Article
    Framework of MM algorithm

    Fig. 3. Framework of MM algorithm

    Download full size
    View in Article
    Plot of ψ(aihi)

    Fig. 4. Plot of ψ(aihi)

    Download full size
    View in Article
    Flow chart of sparse mixture ICP algorithm

    Fig. 5. Flow chart of sparse mixture ICP algorithm

    Download full size
    View in Article
    Initialization of Stanford bunny point-sets with different overlap rates， and the registration results of SM-ICP， Robust Tr-ICP， S-ICP and ICP on these dataset respectively

    Fig. 6. Initialization of Stanford bunny point-sets with different overlap rates， and the registration results of SM-ICP， Robust Tr-ICP， S-ICP and ICP on these dataset respectively

    Download full size
    View in Article
    Plot of the compared registration methods on Stanford bunny point-sets registration

    Fig. 7. Plot of the compared registration methods on Stanford bunny point-sets registration

    Download full size
    View in Article
    Plot of four compared algorithms on the Stanford data （Bun000 and Bun045） with different level of down-sampled points

    Fig. 8. Plot of four compared algorithms on the Stanford data （Bun000 and Bun045） with different level of down-sampled points

    Download full size
    View in Article
    Distance histograms of four comparison algorithms on the registration results of Bun180 and Bun90 point-sets

    Fig. 9. Distance histograms of four comparison algorithms on the registration results of Bun180 and Bun90 point-sets

    Download full size
    View in Article
    Registration results of SM-ICP on the dataset， including Coati， Buddha， Dragon and Stage

    Fig. 10. Registration results of SM-ICP on the dataset， including Coati， Buddha， Dragon and Stage

    Download full size
    View in Article
    Registration results of SM-ICP on the Coati point-sets with different level of noises （N=1 000~5 000）

    Fig. 11. Registration results of SM-ICP on the Coati point-sets with different level of noises （N=1 000~5 000）

    Download full size
    View in Article
    Plot of four compared algorithms on the Coati point-sets registration with different level of noise points

    Fig. 12. Plot of four compared algorithms on the Coati point-sets registration with different level of noise points

    Download full size
    View in Article
    Initialization of engine blade point-sets

    Fig. 13. Initialization of engine blade point-sets

    Download full size
    View in Article
    Registration results of engine blade point-sets for four comparison algorithms.

    Fig. 14. Registration results of engine blade point-sets for four comparison algorithms.

    Download full size
    View in Article
    Vertical view of registration results for SM-ICP and ICP

    Fig. 15. Vertical view of registration results for SM-ICP and ICP

    Download full size
    View in Article

    • Table 1. Both of the errors and time for the four comparison algorithms in the case of Stanford bunny point-sets.

      View table
      View in Article

      Table 1. Both of the errors and time for the four comparison algorithms in the case of Stanford bunny point-sets.

      MethodsOverlap rate
      50%60%75%90%
      SM-ICPTime/s4.72.82.32.1
      Error/m2.04×10-47.75×10-61.42×10-51.14×10-5
      Robust Tr-ICPTime/s0.300.290.120.15
      Error/m1.70×10-37.95×10-61.47×10-51.18×10-5
      S-ICPTime/s1.380.60.20.68
      Error/m8.96×10-44.49×10-47.34×10-41.39×10-4
      ICPTime/s0.250.190.10.13
      Error/m1.8×10-39.07×10-51.5×10-32.53×10-4

    • Table 2. Both trimmed error and registration time of four compared algorithms on the Stanford data （Bun000 and Bun045） with different level of down-sampled points

      View table
      View in Article

      Table 2. Both trimmed error and registration time of four compared algorithms on the Stanford data （Bun000 and Bun045） with different level of down-sampled points

      MethodThe number of down-sampled points
      2 0004 0006 0008 0001 000012 000
      SM-ICPTime/s2.103.955.166.508.158.53
      Error/m1.14×10-51.14×10-51.14×10-51.14×10-51.14×10-51.14×10-5
      Robust Tr-ICPTime/s0.170.310.430.570.730.85
      Error/m1.14×10-51.14×10-51.14×10-51.14×10-51.14×10-51.14×10-5
      S-ICPTime/s0.480.412.152.182.353.37
      Error/m1.39×10-41.40×10-47.60×10-37.00×10-31.90×10-31.70×10-3
      ICPTime/s0.310.270.340.500.540.68
      Error/m2.80×10-42.23×10-42.09×10-41.93×10-41.77×10-41.72×10-4

    • Table 3. Both trimmed error and registration time of four compared algorithms on Coati， Buddha， Dragon and Stage pint-sets registration

      View table
      View in Article

      Table 3. Both trimmed error and registration time of four compared algorithms on Coati， Buddha， Dragon and Stage pint-sets registration

      MethodsDataset
      CoatiBuddhaDragonStage
      SM-ICPTime/s0.922.623.631.33
      Error/m7.50×10-61.93×10-57.70×10-69.45×10-6
      Robust Tr-ICPTime/s0.090.310.660.36
      Error/m7.70×10-61.93×10-67.79×10-61.14×10-5
      S-ICPTime/s0.050.320.750.57
      Error/m1.25×10-53.63×10-58.00×10-41.97×10-5
      ICPTime/s0.040.250.350.318
      Error/m1.69×10-52.09×10-47.69×10-31.14×10-4

    • Table 4. Both trimmed error and registration time of four compared algorithms on the Coati point-sets with different level of noise points

      View table
      View in Article

      Table 4. Both trimmed error and registration time of four compared algorithms on the Coati point-sets with different level of noise points

      MethodsThe number of noises（The number of down-sampling）
      1 000（1 400）2 000（2 700）3 000（4 500）4 000（9 400）5 000（11 550）
      SM-ICPTime/s1.47.0510.455.9514.87
      Error/m4.90×10-66.82×10-53.03×10-58.39×10-51.33×10-4
      Robust Tr-ICPTime/s0.170.070.350.871.69
      Error/m4.98×10-45.52×10-41.89×10-42.67×10-42.13×10-4
      S-ICPTime/s0.090.120.130.421.05
      Error/m4.92×10-45.66×10-45.56×10-42.77×10-42.72×10-4
      ICPTime/s0.040.280.190.621.13
      Error/m5.13×10-44.17×10-43.81×10-42.56×10-43.66×10-4
    Tools

    Get Citation

    Copy Citation Text

    Yue-sheng LIU, Xin-du CHEN, Lei WU, Yun-bao HUANG, Hai yan LI. Sparse mixture iterative closest point registration[J]. Optics and Precision Engineering, 2021, 29(9): 2255

    Download Citation

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

    Category: Information Sciences

    Received: Mar. 23, 2021

    Accepted: --

    Published Online: Nov. 22, 2021

    The Author Email: Yue-sheng LIU (2249791454@qq.com)

    DOI:10.37188/OPE.20212909.2255

    Topics

    laser devices and laser physics
    Lasers and Laser Optics
    Laser physics
    laser manufacturing
    Instrumentation, Measurement and Metrology