• Infrared and Laser Engineering
  • Vol. 50, Issue 12, 20210088 (2021)
Xin Li, Site Mo*, Hua Huang, and Shiji Yang
Author Affiliations
  • College of Electrical Engineering, Sichuan University, Chengdu 610065, China
  • show less
    DOI: 10.3788/IRLA20210088 Cite this Article
    Xin Li, Site Mo, Hua Huang, Shiji Yang. Multi-source point cloud registration method based on automatically calculating overlap[J]. Infrared and Laser Engineering, 2021, 50(12): 20210088 Copy Citation Text show less
    (a) Ideal corresponding points; (b) Nearest distance corresponding points; (c) Convergence curves of algorithms; (d) Distance curves of corresponding points; (e) Slopes between points on the distance curve and the original point; (f) Slope curves of different points
    Fig. 1. (a) Ideal corresponding points; (b) Nearest distance corresponding points; (c) Convergence curves of algorithms; (d) Distance curves of corresponding points; (e) Slopes between points on the distance curve and the original point; (f) Slope curves of different points
    Registration results for Bunny
    Fig. 2. Registration results for Bunny
    Registration results for Hippo
    Fig. 3. Registration results for Hippo
    Registration results for TBunny
    Fig. 4. Registration results for TBunny
    Registration results after adding noise to Bunny
    Fig. 5. Registration results after adding noise to Bunny
    Point cloud of Maoxian landslide. (a) Image reconstruction point cloud; (b) Laser scanning point cloud
    Fig. 6. Point cloud of Maoxian landslide. (a) Image reconstruction point cloud; (b) Laser scanning point cloud
    Registration results of Maoxian landslide point cloud
    Fig. 7. Registration results of Maoxian landslide point cloud
    Point cloudAlgorithmMSETime/s
    Target:Bunny0°(40 097) Source:Bunny45°(40 256) ICP0.108 1604.197
    TrICP0.030 6594.215
    3D-NDT0.032 50763.264
    4PCS0.167 86220.493
    CFB-ICP0.030 6262.111
    Target:Hippo1(30 519) Source:Hippo2(21 935) ICP0.975 6533.280
    TrICP0.807 31912.824
    3D-NDT0.362 541206.879
    4PCS0.517 49323.725
    CFB-ICP0.163 1004.613
    Target:TBunny0°(31 327) Source:Bunny45°(40 256) ICP0.911 2773.092
    TrICP0.027 2145.266
    3D-NDT0.029 57456.107
    4PCS0.149 03120.975
    CFB-ICP0.027 0766.755
    Table 1. MSE and time consumption
    Point cloudAlgorithmMSETime/s
    Target:Noise-Bunny0°(80 194) Source:Noise-Bunny45°(80 512) ICP0.122 02620.157
    TrICP0.065 85559.731
    3D-NDT0.151 467310.675
    4PCS0.157 33950.478
    CFB-ICP0.065 80644.103
    Table 2. MSE and time consumption after adding noise to Bunny
    Point cloudAlgorithmMSETime/s
    Target: Laser scanning point cloud of Maoxian landslide (680 764) Source: Image reconstruction point cloud of Maoxian landslide (2 497 335) ICP36.349 60017.755
    TrICP1.683 42040.739
    3D-NDT2.284 38194.194
    4PCS12.335 44058.285
    CFB-ICP1.678 91013.299
    Table 3. Results comparison of registration for Maoxian landslide
    Xin Li, Site Mo, Hua Huang, Shiji Yang. Multi-source point cloud registration method based on automatically calculating overlap[J]. Infrared and Laser Engineering, 2021, 50(12): 20210088
    Download Citation