基于聚类算法的边缘点集连接方法

Curve reconstruction from scattered plane points using cruster algorithm

针对Lidar点云中轮廓点的连接问题,提出基于聚类(cruster)曲线重建原理进行点云边界连接的方法.首先将三维点云投影到二维平面上;对点云进行细化和约简处理;然后,构建约简后点云的Voronoi图,Voronoi图的顶点近似位于约简点云的中轴上;将中轴点与约简后点云合并在一起,再次进行Voronoi图划分;将新的Voronoi图中顶点包含中轴点的边删除,得到约简后二维点云的边缘轮廓,即简点云的Crust;最后将连接关系映射到三维点云,得到三维轮廓点的连接关系.实验表明,该方法简单,连接结果合理.

For 3D Lidar points, how to connect the edge points should be a problem due to that the spatial relationship among points are more complex. In light of this problem, the cruster algorithm is introduced to reconstruct a curve for 2D points by estimating the middle axis of original points and performing Delau- nay triangulation for all original points and the vertices of middle axis. As the original cruster algorithm only fits the point sets sampled from single smooth curve; when the point sets have certain thickness or uneven thickness, the original cruster algorithm can not work well. However, this is original situation. Then, this paper is to perform thinning and simplification after projecting the 3D points to a 2D plane. Then, the cruster is applied to obtain the edges connecting these simplified points. Finally, the connection relationship between points is mapped to the 3D points. This algorithm is simple and effective.