【摘要】 针对逆向工程基于种子增长区域分割法中种子区域选择和种子增长问题,提出了一种将多尺度理论与区域增长法相结合的点云自动分割方法。采用曲率尺度空间实现多尺度几何分析和特征检测;以多尺度特征检测中大尺度下特征检测作为初始种子区域;基于多尺度特征检测结果之间的相关性,提出按照多尺度特征检测结果进行种子增长;利用转角作为种子是否增长的度量标准;通过特征融合实现特征分割的优化。用多尺度几何分析代替传统单一尺度分析,保证了种子区域选择的稳定性和种子增长方式的可靠性。激光扫描曲面点云分割实例表明:提出的方法可以有效地实现逆向工程特征分割。更多还原

【Abstract】 Aimed at the selection of seed region and seed growing algorithm of region growing for feature segmentation in reverse engineering,an automatic feature segmentation method was developed based on theory of multi-scale and region growing.Curvature scale space was used for the multi scale geometric analysis and feature detection in the proposed method.Based on information correlations among multi-scale feature detection,a new seed growing algorithm was proposed according to multi-scale feature detection.As the homogeneity criteria,relative angle was used to obtain feature segmentation.Feature fusion was presented to obtain an optimization result of feature segmentation.Instead of conventional single scale analysis,multi-scale geometric analysis guarantees the robust selection of seed region and the reliable seed growing algorithm.Practical scanned surface point data indicates that the presented method works effectively on feature segmentation.更多还原