基于曲率阈值和聚类分析的点云关键点提取方法

Point Cloud Key Point Extraction Method Based on Curvature Threshold and Clustering Analysis

针对传统3D扫描在获取点云数据时易受零件粘连和扫描噪声干扰,且数据计算量过大的问题,提出了一种基于曲率阈值和聚类分析的点云关键点提取方法。首先,提出了统计滤波和半径滤波相结合的点云数据预处理方法,避免了噪声点干扰;其次,计算离散点的近似曲率,通过统计曲率分布来选取阈值,进而精简了点云数据;然后,通过聚类算法分析点云密度,来确定特征最为明显的聚类中心点;最后,采用校准算法实现聚类中心和原始点云数据的校准,完成针对零件轮廓的关键点提取。一系列的实验结果表明,该方法无需过多地人为干预参数调整,能快速且高效地实现零件关键点提取,具有较高的准确度。

In order to solve the problem that traditional 3D scanning is vulnerable to the interference of parts sticking and scan-ning noise in extracting point cloud data,and the computation cost is too large,a method of point cloud key point extraction based on curvature threshold and clustering analysis is proposed.First,a method combining statistical filtering and radius filtering is pro-posed to preprocess the point cloud data and avoid the interference of noise points.Subsequently,the approximate curvature of dis-crete points is calculated and the statistical curvature distribution is used to select the threshold.Then,the point cloud density is an-alyzed by clustering algorithm to determine the cluster center,which is also the most obvious point.Finally,the calibration algo-rithm is used to achieve the calibration of the clustering center and the original point cloud data to complete the key point extraction for the part contour.A series of experimental results show that this method can achieve the extraction of key points of parts quickly and efficiently without much human intervention in parameter adjustment,and has high accuracy.