Abstract: Point cloud segmentation is the basis of 3D model retrieval, classification and reconstruction. In order to solve the problems of robustness, over-segmentation and under-segmentation of point cloud segmentation algorithm, a point cloud model segmentation algorithm based on boundary features is proposed. The point cloud model is segmented into weak convex regions. Bhattacharyya distance and modified shape diameter function are used to judge the similarity of adjacent regions and merge similar regions. Through mainstream evaluation methods and experiments, it is proved that most models can achieve good segmentation results.

Key words: point cloud segmentation, boundary point extraction, visibility, Bhattacharyya distance, order relationship analysis method, shape diameter function

摘要： 点云分割是三维模型检索、分类及重建的基础，为解决点云分割算法存在鲁棒性差、过分割和欠分割问题，提出一种基于边界特征的点云模型分割算法。将点云模型过分割为弱凸区域，利用巴氏距离判断相邻区域的相似性进行区域合并，采用改进的形状直径函数进行最终合并。由主流评价方法及实验证明，大多数模型可以取得良好的分割效果。

关键词: 点云分割, 边界点提取, 可见度, 巴氏距离, 序关系分析法, 形状直径函数