一种改进的超体素与区域生长点云分割方法

An improved method for segmentation of supervoxel and regional growing point clouds

点云分割作为识别地理场景的空间特征、探索和记录空间信息的关键处理步骤,其分割精度直接影响后续三维场景重建、地物特征提取等应用的效果。针对传统区域生长点云分割算法的不稳定等问题,本文结合超体素和区域生长算法对点云数据进行分割,并利用点云自身的色彩信息进一步改进分割结果。试验结果表明,相较于传统的区域生长和已有的结合超体素与区域生长的分割算法,本文算法对点云数据分割的效果更好,且其精确率与召回率均有提高。

Point cloud segmentation is a key processing step for identifying spatial features of geographic scenes,exploring and recording spatial information,and its segmentation accuracy directly affects the effects of subsequent 3 D scene reconstruction and feature extraction.Aiming at the instability of traditional region-growing point cloud segmentation algorithms,this paper combines supervoxels and region-growing algorithms to segment point cloud data,and uses the color information of the point cloud itself to further improve the segmentation results.The experimental results show that compared to the traditional region growing and existing segmentation algorithms,combining supervoxels and region growing algorithm proposed in this paper has better effect on point cloud data segmentation,and its accuracy and recall rates are both improved.