基于PCA和三元组的非封闭点云轮廓特征提取方法研究  

Research on contour feature extraction method of unclosed point cloud based on PCA and triplet

作　　者：吴腾飞 王挺[2,3] 邵士亮[2,3] 尚哲 窦鑫可 WU Tengfei;WANG Ting;SHAO Shiliang;SHANG Zhe;DOU Xinke(School of Automation and Electrical Engineering,Shenyang Ligong University,Shenyang 110159;State Key Laboratory of Robotics,Shenyang Institute of Automation,Chinese Academy of Sciences,Shenyang 110016;Institutes for Robotics and Intelligent Manufacturing,Chinese Academy of Sciences,Shenyang 110169;School of Mechanical Engineering and Automation,Northeastern University,Shenyang 110819;School of Information Engineering,Shenyang University of Chemical Technology,Shenyang 110142)

机构地区：[1]沈阳理工大学自动化与电气工程学院,沈阳110159 [2]中国科学院沈阳自动化研究所机器人学国家重点实验室,沈阳110016 [3]中国科学院机器人与智能制造创新研究院,沈阳110169 [4]东北大学机械工程与自动化学院,沈阳110819 [5]沈阳化工大学信息工程学院,沈阳110142

出　　处：《高技术通讯》2023年第12期1303-1312,共10页Chinese High Technology Letters

基　　金：国家自然基金联合基金(U20A20201)资助项目。

摘　　要：针对RGB-D相机获取的物体非封闭点云轮廓特征点和特征线提取不完整的问题,提出了一种基于主成分分析法(PCA)和三元组的轮廓边界提取方法。该方法包括2个步骤:特征点检测和特征线提取。在特征点检测中,通过对邻域点集进行主成分分析,利用采样点曲率和法向夹角双参数判定棱线特征点;根据边界点局部邻域在第二主成分上的分布特点,提出了一种边界检测算子,用于外边界特征点检测。最后,采用三元组聚类方法进行特征线提取。实验结果表明,该方法可以有效地检测非封闭点云模型的轮廓特征点,并分离和提取出目标轮廓曲线,且具有较高的精度。Aiming at the problem of incomplete extraction of the feature points and feature lines from the unclosed point cloud outline of an object acquired by an RGB-D camera,this paper proposes a contour boundary extraction method based on principal component analysis(PCA)and triplets,which consists of two stages:feature point detection and feature line extraction.For feature point detection,via principal component analysis of the set of local neighborhood points,the two parameters of the sample point curvature and the angle between the normal vector are used to determine the ridge-line feature point;depending on the distributional characteristics of the local neighborhood of the limit point on the second principal component,a boundary detection operator is proposed to detect the outer boundary feature points.Lastly,the triplet clustering method is employed for feature extraction.The experimental results show that this method can efficiently detect contour feature points from the unclosed point cloud model,as well as separate and extract the target contour plot with high precision.

关 键 词：非封闭点云 主成分分析(PCA) 特征点检测 特征线提取 三元组 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP242[自动化与计算机技术—计算机科学与技术]