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

Laser & Optoelectronics Progress, Volume. 57, Issue 4, 041008(2020)

Automatic Point Cloud Registration Based on Voxel Downsampling and Key Point Extraction

Bin Zhang1 and Chuanbing Xiong2、*
Author Affiliations
  • 1School of Computer Science, Minnan Normal University, Zhangzhou, Fujian, 363000, China
  • 2School of Physics and Information Engineering, Minnan Normal University, Zhangzhou, Fujian, 363000, China
    • show less
    AbstractGet PDF(in Chinese)
    Figures&Tables (11)
    Equations (0)
    References (16)
    Cited By (11)
    Tools
    Paper Information
    Topics
    Figures & Tables(11)
    Rough registration algorithm flow chart

    Fig. 1. Rough registration algorithm flow chart

    Download full size
    Normal vectors in different regions. (a) Flat region; (b) undulating region

    Fig. 2. Normal vectors in different regions. (a) Flat region; (b) undulating region

    Download full size
    Original point cloud data. (a) Bunny45 original point cloud; (b) dragon00 original point cloud

    Fig. 3. Original point cloud data. (a) Bunny45 original point cloud; (b) dragon00 original point cloud

    Download full size
    Iterative estimation of voxel side length L

    Fig. 4. Iterative estimation of voxel side length L

    Download full size
    View of point cloud after voxel downsampling. (a) Voxel downsampling point cloud based on bunny45 data;(b) voxel downsampling point cloud based on dragon00 data

    Fig. 5. View of point cloud after voxel downsampling. (a) Voxel downsampling point cloud based on bunny45 data;(b) voxel downsampling point cloud based on dragon00 data

    Download full size
    Extraction effect of key points under different algorithms based on bunny45 data and dragon00 data. (a)(d) Algorithm in Ref. [6]; (b)(e) algorithm in Ref. [7]; (c)(f) our algorithm

    Fig. 6. Extraction effect of key points under different algorithms based on bunny45 data and dragon00 data. (a)(d) Algorithm in Ref. [6]; (b)(e) algorithm in Ref. [7]; (c)(f) our algorithm

    Download full size
    Matching results of corresponding key point pairs. (a)(c) Unstreamlined feature points matching; (b)(d) streamlined feature points matching

    Fig. 7. Matching results of corresponding key point pairs. (a)(c) Unstreamlined feature points matching; (b)(d) streamlined feature points matching

    Download full size
    Rough registration effect of different algorithms. (a)(e) Original point cloud before registration; (b)(f) preliminary rough registration effect of algorithm in Ref. [6]; (c)(g) preliminary rough registration effect of algorithm in Ref. [7]; (d)(h) rough registration effect of our algorithm

    Fig. 8. Rough registration effect of different algorithms. (a)(e) Original point cloud before registration; (b)(f) preliminary rough registration effect of algorithm in Ref. [6]; (c)(g) preliminary rough registration effect of algorithm in Ref. [7]; (d)(h) rough registration effect of our algorithm

    Download full size
    Fine registration effect of rough registration results for different algorithms. (a)(d) Fine registration effect for algorithm in Ref. [6]; (b)(e) fine registration effect for algorithm in Ref. [7]; (c)(f) fine registration effect for our algorithm

    Fig. 9. Fine registration effect of rough registration results for different algorithms. (a)(d) Fine registration effect for algorithm in Ref. [6]; (b)(e) fine registration effect for algorithm in Ref. [7]; (c)(f) fine registration effect for our algorithm

    Download full size

    • Table 1. Comparison of rough registration error and running time of each algorithm

      View table

      Table 1. Comparison of rough registration error and running time of each algorithm

      Raw pointcloud dataError beforeregistration /10-3Algorithm in Ref. [6]Algorithm in Ref. [7]Our algorithm
      MSE /mTime /sMSE /mTime /sMSE /mTime /s
      Bunny00, bunny451.10×10-35.53×10-560.01±1.468.13×10-522.14±0.641.93×10-511.17±0.18
      Dragon00, dragon487.01×10-36.81×10-567.77±1.523.68×10-418.22±0.411.99×10-412.87±0.15

    • Table 2. Comparison of fine registration error and total registration time of rough registration results for each algorithm

      View table

      Table 2. Comparison of fine registration error and total registration time of rough registration results for each algorithm

      Raw point cloud dataFor algorithm in Ref. [6]For algorithm in Ref. [7]For our algorithm
      MSE /(10-6 m)Time /sMSE /(10-6 m)Time /sMSE /(10-6 m)Time /s
      Bunny00, bunny455.0062.25±1.415.0724.98±0.555.0812.91±0.13
      Dragon00, dragon4835.0069.58±1.6388.5018.22±0.4135.0015.08±0.08
    Tools

    Get Citation

    Copy Citation Text

    Bin Zhang, Chuanbing Xiong. Automatic Point Cloud Registration Based on Voxel Downsampling and Key Point Extraction[J]. Laser & Optoelectronics Progress, 2020, 57(4): 041008

    Download Citation

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

    Category: Image Processing

    Received: May. 30, 2019

    Accepted: Jul. 25, 2019

    Published Online: Feb. 20, 2020

    The Author Email: Xiong Chuanbing (1136466643@qq.com)

    DOI:10.3788/LOP57.041008

    Topics

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