基于区域增长方法的点云分割

Point Cloud Segmentation Based on Regional Growth Method

点云分割作为现在机器视觉中的热点问题是点云分类、识别以及三维重建的基础,分割结果对后续的场景分析有着重要作用。本文通过对点云分割展开研究,对三维点云数据进行体素化处理得到体素云,然后在体素空间内进行网格化处理来筛选种子体素,再利用聚类算法形成超体素,完成过分割的处理过程。根据区域增长算法原理,对过分割得到的超体素数据进行平面拟合,实现点云数据的分割。实验结果表明本文的分割算法可以提高处理效率,使计算量大大减少,并且使最后的分割结果更加精确。

As a hot issue in machine vision,point cloud segmentation is the basis of point cloud classification,recognition and 3 D reconstruction.The segmentation results play an important role in the subsequent scene analysis.In this paper,point cloud segmentation is studied,and voxel cloud is obtained by voxel processing of three-dimensional point cloud data.Then,seed voxel is screened by grid processing in voxel space,and supervoxel is formed by clustering algorithm to complete the process of over-segmentation.According to the principle of region growth algorithm,plane fitting is carried out for the hypervolume prime data obtained from over-segmentation to realize the segmentation of point cloud data.The experimental results show that the proposed segmentation algorithm can improve the processing efficiency,greatly reduce the computation,and make the final segmentation result more accurate.