测绘通报 ›› 2024, Vol. 0 ›› Issue (4): 1-5.doi: 10.13474/j.cnki.11-2246.2024.0401

• 学术研究 •    

改进ICP算法的激光雷达点云配准

许哲1,2, 董林啸1, 吴家跃1   

  1. 1. 上海海洋大学工程学院, 上海 201306;
    2. 上海海洋可再生能源工程技术研究中心, 上海 201306
  • 收稿日期:2023-09-01 发布日期:2024-04-29
  • 通讯作者: 董林啸。E-mail:linxiao982022@163.com
  • 作者简介:许哲(1970—),男,博士,副教授,主要研究方向为机器人控制、图像识别技术。E-mail:946621972@qq.com
  • 基金资助:
    上海市联盟计划(D-8006-05-0031);上海市科学技术委员会资助项目(19DZ2254800)

LiDAR point cloud registration with improved ICP algorithm

XU Zhe1,2, DONG Linxiao1, WU Jiayue1   

  1. 1. College of Engineering Science and Technology, Shanghai Ocean University, Shanghai 201306, China;
    2. Shanghai Engineering Research Center of Marine Renewable Energy, Shanghai 201306, China
  • Received:2023-09-01 Published:2024-04-29

摘要: 针对传统ICP算法在激光雷达目标点云配准中存在匹配时间长,以及受初值影响导致该算法应用在无人车SLAM技术中容易存在定位精度不高和稳健性较差的问题,本文提出了一种结合KD-tree算法的NDT-ICP算法。首先,通过Voxel Grid滤波对激光雷达获取的点云数据进行预处理,利用平面拟合参数的方法去除地面点云;然后,利用NDT算法进行点云粗匹配,缩短目标点云与待匹配点云距离;最后,通过KD-tree邻近搜索法提高对应点查找速度,并通过优化收敛阈值,完成ICP算法的精匹配。试验结果表明,本文提出的改进算法相比于NDT算法和ICP算法,在点云配准速度和精度上有明显提高,且在地图构建上精度和稳健性更好。

关键词: 无人车, 点云配准, ICP算法, NDT算法, 激光SLAM

Abstract: The traditional ICP algorithm has long matching time and is affected by initial values in LiDAR target point cloud matching, which leads to low positioning accuracy and poor robustness when applied to unmanned vehicle SLAM technology. Proposes an NDT-ICP algorithm that combines the KD-tree algorithm. Firstly, voxel grid filtering is used to preprocess the point cloud data obtained from LiDAR, and the method of plane fitting parameters is used to remove point cloud of ground. Secondly, use NDT algorithm for point cloud coarse matching to shorten the distance between the target point cloud and the point cloud to be matched. Finally, the KD-tree proximity search method is used to improve the speed of corresponding point search, and the precise matching of the ICP algorithm is completed by optimizing the convergence threshold. Through experiments, it has been shown that the improved algorithm proposed in this article has significantly improved speed and accuracy in point cloud matching compared to NDT and ICP algorithms, and has better accuracy and robustness in map construction.

Key words: unmanned vehicle, point cloud registration, ICP algorithm, NDT algorithm, laser SLAM

中图分类号: