Chinese Optics, Volume. 17, Issue 4, 875(2024)
An improved point cloud registration method based on the point-by-point forward method
Fig. 2. Schematic diagram of affine invariant 4-point set extraction
Fig. 3. Schematic diagram of bidirectional Kd-tree point pair search
Fig. 7. Registration results of Stanford point cloud data with Gaussian noise of
Fig. 8. Registration error of each algorithm with different Gaussian noises
Fig. 9. Registration consuming times of each algorithm with different Gaussian noises
Fig. 10. Registration results of object point cloud data in indoor scene
Table 1. RGB-D camera’s parameters
View tableView in ArticleTable 1. RGB-D camera’s parameters
参数名称 数值 工作范围 0.6~8 m 深度 精度 l m: ±3 mm 视场角（FOV) H58.4 × V45.5° 分辨率@帧率 640 × 480@30 fps 视场角（FOV) H66.1° × 740.2° RGB 分辨率@帧率 640 × 480@30 fps UVC 支持
Table 2. Quantitative analysis results of Stanford point cloud data registration
View tableView in ArticleTable 2. Quantitative analysis results of Stanford point cloud data registration
Model SAC-IA+ICP NDT+ICP Ours RMSE/mm Time/s RMSE/mm Time/s RMSE/mm Time/s Armadillo 0.0288 3.21 0.0168 2.19 0.00730 0.437 Dragon 0.0363 3.77 0.0218 3.25 0.0125 0.528
Table 3. Quantitative analysis results of object point cloud data registration in indoor scene
View tableView in ArticleTable 3. Quantitative analysis results of object point cloud data registration in indoor scene
模型 源点云数/个 源点云特征点数/个 目标点云数/个 目标点云特征点数/个 RMSE/mm 平均误差/mm 耗时/s 平均耗时/s Chair 8549 3702 10431 4847 0.0686 0.544 Kettle 3713 1359 3879 1449 0.0633 0.0742 0.487 0.572 Mannequin 45292 16098 46507 16462 0.0907 0.686
Get Citation
Copy Citation Text
Mao-yue LI, Sheng-bo XU, Ling-qiang MENG, Zhi-cheng LIU. An improved point cloud registration method based on the point-by-point forward method[J]. Chinese Optics, 2024, 17(4): 875
Received: Sep. 22, 2023
Accepted: --
Published Online: Aug. 9, 2024
The Author Email:
