Six dimensional clustering segmentation of color point cloud

Abstract

This paper focuses on the clustering segmentation of 3D color point cloud. We extend the mean shift algorithm to the 3D xyz space, and what's more, we also consider the rgb color information, so the 6 dimensional data is adopted in the algorithm. The cluster center converges to the joint position of the local maximum density and the minimum gradient change of color, so our clustering segmentation not only considers the local geometrical features, but also utilizes the color information. The experiments show that our segmentation has better region consistency and has clear segmenting border in different color neighbors.