• Optics and Precision Engineering
  • Vol. 32, Issue 10, 1606 (2024)
Jinru LI1,2, Jin WANG3,*, Songtao GUO3, and Hongyan SUO1
Author Affiliations
  • 1Shanxi Coal Geological Survey and Mapping Institute Co., Ltd,Jinzhong030600,China
  • 2School of Surveying and Spatial Information, Shandong University of Science and Technology, Qingdao66590,China
  • 3School of Geospatial Information, University of Information Engineering, Strategic Support Force of the People's Liberation Army of China,Zhengzhou450001,China
  • show less
    DOI: 10.37188/OPE.20243210.1606 Cite this Article
    Jinru LI, Jin WANG, Songtao GUO, Hongyan SUO. Point cloud matching algorithm based on adaptive local neighborhood conditions[J]. Optics and Precision Engineering, 2024, 32(10): 1606 Copy Citation Text show less
    Feature points extraction process
    Fig. 1. Feature points extraction process
    Neighborhood of point pi
    Fig. 2. Neighborhood of point pi
    Distribution of normal vectors in different regions
    Fig. 3. Distribution of normal vectors in different regions
    Registration process
    Fig. 4. Registration process
    Rang of feature point pi and its neighborhood points estimated
    Fig. 5. Rang of feature point pi and its neighborhood points estimated
    uvw Local coordinate system
    Fig. 6. uvw Local coordinate system
    Influence of different thresholds on registration
    Fig. 7. Influence of different thresholds on registration
    Influence of different extraction methods on registration
    Fig. 8. Influence of different extraction methods on registration
    Registration effects of Bunny point cloud
    Fig. 9. Registration effects of Bunny point cloud
    Registration effects of Dragon point cloud
    Fig. 10. Registration effects of Dragon point cloud
    Point cloud date of teaching building J1
    Fig. 11. Point cloud date of teaching building J1
    Number of extracted feature points and registration error under different threshold conditions
    Fig. 12. Number of extracted feature points and registration error under different threshold conditions
    Proposed algorithm registration results of school gate(Blue is the point cloud to be configured; Red is the target point cloud;)
    Fig. 13. Proposed algorithm registration results of school gate(Blue is the point cloud to be configured; Red is the target point cloud;)
    Different algorithm registration results of school gate
    Fig. 14. Different algorithm registration results of school gate
    噪声传统ICP算法ISS+SAC-IA+ICPK-4PCS+ICP本文算法
    RMSE/mmTime/sRMSE/mmTime/sRMSE/mmTime/sRMSE/mmTime /s
    σ=00.23232.470.15225.850.33923.350. 03518.01
    σ=0.0010.48745.210.36331.540.52532.900. 04022.35
    σ=0.021.12759.640.52640.790.89440.530. 04926.07
    Table 1. Comparison of registration results of different algorithms for Bunny point cloud
    数据缺失传统ICP算法ISS+SAC-IA+ICPK-4PCS+ICP本文算法
    RMSE/mmTime/sRMSE/mmTime/sRMSE/mmTime/sRMSE/mmTime/s
    0%0.20176.960.07643.620.13985.130.03830.25
    20%0.45880.700.20139.160.24222.090.03628.12
    50%1.52436.861.31247.572.5676.710.04025.85
    Table 2. Comparison of registration results of different algorithms for Dragon point cloud
    实验数据传统ICP算法K-4PCS+ICP配准ISS+SAC-IA+ICP本文算法配准
    RMSE/cmTime/sRMSE/cmTime/sRMSE/cmTime/sRMSE/cmTime/s
    校门80.175269.3228.24251.40927.968282.1615.12172.52
    Table 3. Comparison of different algorithmic alignment assessment metrics
    Jinru LI, Jin WANG, Songtao GUO, Hongyan SUO. Point cloud matching algorithm based on adaptive local neighborhood conditions[J]. Optics and Precision Engineering, 2024, 32(10): 1606
    Download Citation