基于改进ICP的帧间匹配算法研究  

Research on inter⁃frame matching algorithm based on improved ICP

作　　者：刘贺伟 卿兆波 黄加俊 LIU Hewei;QING Zhaobo;HUANG Jiajun(College of Mechanical and Electrical Engineering,China Jiliang University,Hangzhou 310000,China)

机构地区：[1]中国计量大学机电工程学院,浙江杭州310000

出　　处：《现代电子技术》2023年第8期73-78,共6页Modern Electronics Technique

摘　　要：ICP算法在每一次迭代时均会对点云数据中每个点进行选择,直到得出与目标点最近点,再计算得出位移向量和旋转矩阵,存在迭代计算时间较长、精度较低的问题。为解决该问题,文中提出一种以欧氏距离为匹配条件,利用法向量信息找出匹配点,同时利用实际曲面信息的特征过滤错误匹配点的算法。所提算法利用中心重合法来降低前后两帧激光点云数据的平移距离,利用KD-Tree算法将各个激光点按照坐标分类,最后利用激光点的法线和曲率淘汰错误点,提高匹配精度。在Anaconda环境下进行仿真实验,结果表明,在点云数目为363,727,1 400时,采用改进ICP算法所用时间分别是原ICP算法的39%,41%,56%,均方根误差分别控制在原ICP算法的0.68%,18.48%,18.04%。改进后的ICP算法匹配效果得到显著提升,误差减小,迭代时间明显缩短。In order to solve the problem that current ICP algorithm has long iterative calculation time and low accuracy becauce it selects each point in the point cloud data in each iteration until the point closest to the target point is obtained,and then the displacement vector and rotation matrix are calculated,an algorithm that takes the Euclidean distance as the matching condition,uses the normal vector information to find the matching points,and utilises the features of the actual curved surface information to filter the wrong matching points is proposed.In this algorithm,the center coincidence method is used to reduce the translation distance of laser point cloud data of front frame and rear frame,the KD⁃Tree algorithm is used to classify all the laser points according to their coordinates,and the normal and curvature of the laser point are utilized to eliminate the wrong points to improve the matching accuracy.The simulation experiment was carried out in Anaconda environment.The results show that when the number of point clouds is 363,727 and 1400,the time taken by the improved ICP algorithm is 39%,41%and 56%of the original ICP time respectively,and its root mean square error can be controlled within 0.68%,18.48%and 18.04%of the original ICP algorithm respectively.The experimental results show that the matching effect of the improved ICP algorithm is improved obviously,the error is reduced,and the iteration time is significantly shortened.

关 键 词：帧间匹配 改进ICP算法 匹配点过滤 激光点分类 错误点淘汰 仿真实验 

分 类 号：TN919-34[电子电信—通信与信息系统] TP391.41[电子电信—信息与通信工程]