基于模糊极大似然估计聚类的点云数据分块

Point-cloud Data Segmentation Based on Fuzzy Maximum Likelihood Estimate Clustering

对散乱点云数据采用微切平面法进行法矢估计,对法矢方向进行全局协调性调整。采用稳定性较好的二次曲面拟合法估算点云数据的高斯曲率和平均曲率。将点的坐标、法矢和曲率合并为八维特征向量,通过模糊极大似然估计聚类技术,将具有类似几何特征的向量聚为一类,从而实现点云数据的分块。实验证明该方法有效。

The normal vector at each point is estimated by finding a least square fitted plane of that point and its neighbors, and all of the estimated normal vectors are globally adjusted. Gaussian and mean curvatures at each point are estimated by quadric fitting. Eight dimensional feature vectors consisting of coordinates, normal vector, mean curvature and Gaussian curvature are taken as input feature vectors. By applying the fuzzy maximum likelihood estimate clustering method, the segmentation is implemented. Experimental results show the validity of the method.