Light Detection and Ranging (LiDAR) is gradually developing towards low-cost and high reliability, a variety of highperformance products using different techniques have been derived, which have been widely used in pedestrian detection. In this work, we choose a repetitive scanning LiDAR and a non-repetitive scanning LiDAR to compare the distribution of pedestrian point cloud at different distances. In addition, a pedestrian detection algorithm based on density clustering is designed to compare the detection effects of the two kinds of devices, which provide data support for the research of smart security, V2X (vehicle-to-everything), and autonomous driving. The experiment results show that Livox Horizon has the ability of capturing pedestrian cross-section point cloud with higher completeness and density than Ouster OS1- 64 as integration time increases. Moreover, Horizon and OS1-64 have basically the same detection effect on closedistance dynamic pedestrian, and OS1-64 performs better when detecting pedestrian at 40m. By means of growing integration time, Horizon greatly enhances the ability of detecting long-distance pedestrian.
|