Journal of Applied Optics, Volume. 44, Issue 2, 330(2023)
Point cloud registration algorithm based on 3D shape context features
Figures & Tables(16)
Fig. 2. Comparison of point cloud before and after voxel filtering
Fig. 3. Comparison results of three commonly-used point cloud feature extraction methods
Table 1. The number of feature points extracted by different methods
View tableView in ArticleTable 1. The number of feature points extracted by different methods
特征点提取方法 特征点数 时间/s ISS算法 66 0.403 3D-harris算法 35 0.692 3D-sift算法 52 1.265
Table 2. The number of feature points extracted by ISS algorithm under different search radii
View tableView in ArticleTable 2. The number of feature points extracted by ISS algorithm under different search radii
搜索半径/m 0.03 0.04 0.05 0.06 0.07 源点云 1159 591 444 388 363 目标点云 1381 707 582 509 486
Table 3. Other main parameters of proposed algorithm
View tableView in ArticleTable 3. Other main parameters of proposed algorithm
参数 数值 ISS L2/L1的阈值 0.65 ISS L3/L2的阈值 0.5 3DSC球面最小半径/m 0.02 3DSC领域点半径/m 0.1 3DSC密度计算阈值 0.02
Table 4. Number of point clouds and matching of feature points
View tableView in ArticleTable 4. Number of point clouds and matching of feature points
数据集 点云原始数量 下采样后的点云数量 特征点个数 粗匹配前的匹配点对 粗匹配后的匹配点对 匹配成功率/% bunny 34834 29962 19 14 12 85.7 34825 29187 21 fr1_desk 189053 158565 444 72 62 86.1 186134 156961 582 fr2_desk 159187 135065 119 16 14 87.5 146533 124076 147 fr2_xyz 1011581 160174 1060 41 35 85.3 1075288 237301 1381
Table 5. Evaluation of registration effect of bunny
View tableView in ArticleTable 5. Evaluation of registration effect of bunny
配准方法 时间/s 均方根误差/m 本文算法 0.041 4.70×10−7 SAC-IA+ICP算法 1.023 7.09×10−4 ISS+3DSC+ndt算法 0.094 5.46×10−5
Table 6. Evaluation of registration effect of fr1_desk
View tableView in ArticleTable 6. Evaluation of registration effect of fr1_desk
配准方法 时间/s 均方根误差/m 本文算法 0.229 2.00×10−4 SAC-IA+ICP算法 11.403 4.54×10−4 ISS+3DSC+ndt算法 0.231 3.05×10−4
Table 7. Evaluation of registration effect of fr2_desk
View tableView in ArticleTable 7. Evaluation of registration effect of fr2_desk
配准方法 时间/s 均方根误差/m 本文算法 0.063 2.31×10−4 SAC-IA+ICP算法 8.223 4.50×10−4 ISS+3DSC+ndt算法 0.074 3.02×10−4
Table 8. Evaluation of registration effect of fr2_xyz
View tableView in ArticleTable 8. Evaluation of registration effect of fr2_xyz
配准方法 时间/s 均方根误差/m 本文算法 2.807 1.68×10−5 SAC-IA+ICP算法 14.976 4.91×10−4 ISS+3DSC+ndt算法 2.887 2.99×10−5
Tools
Get Citation
Copy Citation Text
Zixiang ZHOU, Dandan HUANG, Zhi LIU. Point cloud registration algorithm based on 3D shape context features[J]. Journal of Applied Optics, 2023, 44(2): 330
Paper Information
Category: Research Articles
Received: May. 9, 2022
Accepted: --
Published Online: Apr. 14, 2023
The Author Email: HUANG Dandan (hdd@cust.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20