The content of this article comes from the "Surveying and Mapping Bulletin" No. 4 in 2024, drawing review number: GS Jing (2024) No. 0563
LiDAR point cloud registration with improved ICP algorithm
XU Zhe1,2, DONG Linxiao1, WU Jiayue1
1. School of Engineering, Shanghai Ocean University, Shanghai 201306, China; 2. Shanghai Marine Renewable Energy Engineering Technology Research Center, Shanghai 201306
Funds: Shanghai Alliance Program (D-8006-05-0031); The Shanghai Municipal Science and Technology Commission(19DZ2254800)
Keywords: unmanned vehicle, point cloud registration, ICP algorithm, NDT algorithm, laser SLAM
Citation format: Xu Zhe, Dong Linxiao, Wu Jiayue. LiDAR Point Cloud Registration Based on Improved ICP Algorithm[J]. Bulletin of Surveying and Mapping, 2024(4): 1-5. DOI: 10.13474/j.cnki.11-2246.2024.0401
Abstract:In view of the long matching time of the traditional ICP algorithm in the registration of LiDAR target point cloud, and the problem that the algorithm is prone to low positioning accuracy and poor robustness in the SLAM technology of unmanned vehicles due to the influence of the initial value, this paper proposes an NDT-ICP algorithm combined with KD-tree algorithm. Firstly, the point cloud data obtained by the lidar was preprocessed by Voxel Grid filtering, and the ground point cloud was removed by the method of plane fitting parameters. Then, the NDT algorithm was used to coarsely match the point cloud to shorten the distance between the target point cloud and the point cloud to be matched. Finally, the KD-tree proximity search method was used to improve the corresponding point finding speed, and the precise matching of the ICP algorithm was completed by optimizing the convergence threshold. Experimental results show that compared with the NDT algorithm and the ICP algorithm, the improved algorithm proposed in this paper has a significant improvement in the speed and accuracy of point cloud registration, and has better accuracy and robustness in map construction.
body
About author:XU Zhe (1970—), male, Ph.D., associate professor, main research direction is robot control, image recognition technology. E-mail:[email protected] Corresponding author: Dong Linxiao. E-mail:[email protected]
First trial: Ji Yinxiao review: Song Qifan
Final Judge: Jin Jun
information