Improved color region growing point cloud segmentation algorithm based on octree

Abstract

Aiming at the problems that the traditional color region growing segmentation algorithm has a large amount of computation, slow running speed and is easily affected by noise, this paper proposes an improved color region growing point cloud segmentation algorithm based on octree. The proposed algorithm consists of two segmentation stages from coarse to fine: firstly, an octree-based voxelized representation of the input point cloud is performed, and a traditional region growing algorithm segmentation step is performed to extract the main (coarse) parts. Then, the region growth of boundary points is performed by replacing geometric features with color features to achieve fine segmentation. The experimental results show that this method can not only effectively segment point cloud data, but also solve the problem of instability of traditional color-based region growth segmentation, and improve the accuracy, reliability and running speed of point cloud segmentation.