关键词: 实物表面采样数据, 曲面边界样点识别, 均值漂移, 核密度估计, 动态空间索引

Abstract: For solving the problem that current surface boundary points detection algorithms were difficult to adapt non-uniform distributed sampled data of physical surface,a boundary detection algorithm based on reverse mean shift was proposed.Based on mean shift algorithm,the surface local sample which used by k-nearest neighbors of objective point was extended to the sampled data sparse region of adjacent objective point,and the gain optimization for the surface local sample was realized.The kernel density estimation was applied for gain optimized sample to obtain the corresponding mode point of objective point.The boundary points were detected by comparing the deviation extent between the objective point and its mode point.The experimental results showed that the proposed algorithm could detect the characteristic points of surface trim boundary,public boundary of geometric continuous adjacent surfaces and transitional curved surface with great curvature change,and had good adaptability for the sample data of non-uniform distribution.

Key words: sampled data of physical surface, surface boundary point detection, mean-shift, kernel density estimation, dynamic spatial index