Segmentation Method of Pedestrian Object in Ground Point Cloud Image

Aiming at the problem of incomplete segmentation of target contour information in traditional pedestrian target segmentation methods under different gaits such as walking and jogging, this paper presents a segmentation method for pedestrian objects in ground point cloud image. Firstly, statistical outlier removal combined with the random sample consensus method is proposed to remove outliers and segment all plane point clouds. Then, based on the ground point cloud image, a range maximum search method is proposed to extract the region of interest. Finally, the pass-through filter is used to segment pedestrian targets in the region of interest. In this paper, point cloud data of four gaits of three pedestrians are collected and compared with traditional Euclidean Clustering and Density-Based Spatial Clustering of Applications with Noise (DBSCAN) Clustering. The results show that the average accuracy of pedestrian target segmentation under different gait is 92.96%. The validity and advance of the proposed algorithm are proved.