现代制造工程 ›› 2026, Vol. 552 ›› Issue (9): 68-75.doi: 10.16731/j.cnki.1671-3133.2026.09.008

• CAD/CAE/CAPP/CAM • 上一篇    下一篇

基于多阈值的层次化点云边界提取及修正方法

顾登明1, 李龙1, 许小进1, 王荣2   

  1. 1 国核电站运行服务技术有限公司,上海 200233;
    2 东华大学机械工程学院,上海 201620
  • 收稿日期:2025-08-26 出版日期:2026-09-18 发布日期:2026-10-22
  • 作者简介:顾登明,本科,高级工程师,主要研究方向为核燃料检查。E-mail:gudm@snpsc.com
  • 基金资助:
    国家电力投资集团研发项目(GHYX-HX-2301)

A hierarchical point cloud boundary extraction and correction method based on multiple thresholds

GU Dengming1, LI Long1, XU Xiaojin1, WANG Rong2   

  1. 1 State Nuclear Power Plant Service Co.,Ltd.,Shanghai 200233,China;
    2 College of Mechanical Engineering,Donghua University,Shanghai 201620,China
  • Received:2025-08-26 Online:2026-09-18 Published:2026-10-22

摘要: 针对现有点云边界提取算法依赖单一判定准则导致检测精度较低的问题,提出了一种基于多阈值的层次化点云边界快速精确提取算法,并针对提取结果进行了修正处理。采用KD-Tree建立点间拓扑结构,将采样点邻域点集质心偏移量与邻域点矢量和的判定准则加权融合,实现对边界特征明显的点云区域的初步识别提取;遍历初步提取边界点集进行近邻搜索,将采样点及其邻域点投影至最小二乘拟合平面上,通过将投影点所构成向量间夹角的最大值设定为阈值的方法进行精确提取;将提取结果划分为特征区域与非特征区域,采用移动最小二乘法分别进行修正处理。通过对多种工件的点云数据集进行试验,证明了该算法可以快速精确地提取出连续且平滑的工件点云边界,且具有良好的稳定性。

关键词: 三维点云, 边界提取, 移动最小二乘法, 点云平滑

Abstract: Aiming at the problem of low detection accuracy caused by the existing point cloud boundary extraction algorithm relying on a single criterion,a fast and accurate extraction algorithm of hierarchical point cloud boundary based on multiple thresholds was proposed, and the extraction results were corrected. KD-Tree was used to establish the topological structure between points. The weighted fusion of the centroid offset of the neighborhood cluster points of the sampling points and the criterion of the sum of the neighborhood point vectors was used to preliminarily extract the point cloud clusters with boundary features. The boundary cluster points were traversed for neighbor search. The sampling points and neighborhood points were projected onto the least squares fitting plane. Accurate extraction was performed by the method of the maximum angle threshold between the vectors formed by the projection points as the threshold value. The extraction results were divided into feature regions and non-feature regions. The moving least square method was used to correct the processing respectively. Through experiments on point cloud data sets of various artifacts,it was proved that the algorithm can extract continuous and smooth workpiece point cloud boundary quickly and accurately,and has good stability.

Key words: 3D point cloud, boundary extraction, moving least squares, point cloud smoothing

中图分类号: 

版权所有 © 《现代制造工程》编辑部 
地址:北京市东城区东四块玉南街28号 邮编:100061 电话:010-67126028 电子信箱:2645173083@qq.com
本系统由北京玛格泰克科技发展有限公司设计开发 技术支持:support@magtech.com.cn