Distance based neighbor correlation for the segmentation

摘要

This paper introduces the segmentation of point cloud with the distance-based connectivity that is originated from the connectivity of local convexity criterion to enhance its accuracy and singularity [1]. The proposed feature is applied to calculate the weighted normal vector and to partition ground and objects respectively through integrating it with other features. The performances of segmentations with the introduced criterion are demonstrated with the labeled simulation data and the real data from 3D LIDAR compared to the original connectivity.