一种结合边缘点和特征点的点云精简方法及其效果评价  

A Point Cloud Reduction Method of Combining Edge Points and Feature Points and its Effect Evaluation

作　　者：杨森 花向红[1,2] 李丞 水浩奇 高金彬 张涧栗 YANG Sen;HUA Xianghong;Li Cheng;SHUI Haoqi;GAO Jinbin;ZHANG Jianli(School of Geodesy and Geomatics,Wuhan University,Wuhan 430079,China;Disaster Monitoring and Prevention Research Center of Wuhan University,Wuhan 430079,China)

机构地区：[1]武汉大学测绘学院,湖北武汉430079 [2]武汉大学灾害监测与防治研究中心,湖北武汉430079

出　　处：《城市勘测》2023年第5期87-92,共6页Urban Geotechnical Investigation & Surveying

基　　金：国家自然科学资金面上项目(42271452)。

摘　　要：针对获得的点云原始数据十分庞大和原始数据中存在大量冗余的问题,提出一种结合边缘点和特征点提取的点云精简方法。该方法首先根据主成分分析(PCA)算法提取点云边缘点;然后,建立点云的K领域关系计算点云的ISS特征点,对点云构造尺度空间求得点云的SIFT3D特征点,从而提取点云特征点;第三,在提取出点云边缘点和点云内部的特征点后,在点云平坦区域可能会存在明显的空洞,采用改进的最远点采样算法对空洞进行点云增补;最后,将边缘点、特征点和最远点采样得到的三类点云合并,设置阈值去除重复点,从而实现点云精简。最终采用精简率、豪斯多夫距离和几何失真指数三个指标评价精简效果,实验结果表明:本文方法不但避免了点云精简后平坦区域的空洞问题,还能很好地保留点云的细节特征,根据提出的几何失真指数评价指标可以得到本文的精简方法优于随机采样法、均匀采样法和法向空间采样法。For the problem that raw point cloud data is very large and has a large amount of redundancy,a point cloud streamlining method combining edge point and feature point extraction is proposed.This method first extracts the edge points according to the principal component analysis(PCA)algorithm.And then,the ISS feature points are calculated by the K domain relationship of the point cloud,and the SIFT3D feature points are obtained by constructing a scale space for the point cloud,so as to extract the point cloud feature points.Thirdly,there may be obvious voids in the flat area of the point cloud after extracting edge points and feature points,the improved farthest point sampling algorithm is used to add point clouds to the voids.Finally,the three types of point clouds obtained from edge points,feature points and farthest point sampling are combined,and a threshold is set to remove duplicate points,so as to realize point cloud streamlining.We use streamlining rate,Hausdorff distance and geometric distortion index to evaluate the streamlining effect.The experimental results show that this method not only avoids the problem of hollowness in the flat area after point cloud streamlining,but also retains the detailed features of the point cloud well,and according to the proposed geometric distortion index evaluation index,we can get that the streamlining method of this paper is better than random sampling method,uniform sampling method and normal space sampling method.

关 键 词：边缘点提取 特征点提取 点云增补与删除 点云精简 精简评价指标 

分 类 号：P225.2[天文地球—大地测量学与测量工程]