基于边特征的点云数据区域分割  

Edge-based segmentation of point cloud data

作　　者：柯映林[1] 单东日[2] 

机构地区：[1]浙江大学机械与能源工程学院,浙江杭州310027 [2]山东轻工业学院机械与电气学院,山东济南250100

出　　处：《浙江大学学报（工学版）》2005年第3期377-380,396,共5页Journal of Zhejiang University：Engineering Science

基　　金：教育部优秀骨干教师基金和教育部博士点专项基金资助项目(9803352).

摘　　要：为了提高反求工程建模的效率,提出了一种基于空间栅格的区域分割方法.该方法采用二次抛物面模型计算散乱数据点的曲率,利用空间栅格结构建立散乱点的拓扑关系,根据栅格中数据点与栅格中心点的相对位置计算栅格曲率以及相邻栅格间的曲率差值,由曲率差函数判别并抽取边特征栅格.通过特征栅格的空间位置与曲面栅格的连通性实现了空间散乱数据的区域分割. 工程应用实例表明: 该方法能够直接处理点云数据而无需构建三角网格,具有运算简单,稳定性高等特点.可有效解决具有曲率突变性的点云数据的区域分割及特征提取问题.To improve the efficiency of reverse modeling, a new grid-based segmentation method not using triangulation of unorganized points was proposed. The method estimated curvature by coordinate transformation method, calculated the enclosure box of a large amount of measured points and subdivided the box into many 3D grids, each 3D grids containing some scattered points. The grid's curvature was set as the maximum principal curvature of the point nearest to the grid center. According to the grid's curvature and location, a special function describing the curvature difference between the grid and its neighbors-was built. Based on this function the edge-grids were detected and the point cloud was partitioned into several regions based on the connectivity of 3D grids. Application results prove that the proposed method can process point cloud directly, and is feasible and efficient in the segmentation of data sampled from objects with edges.

关 键 词：区域分割 特征提取 点云数据 反求工程 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]