LiDAR点群とカメラ画像を複合した移動ロボットからの歩行者追尾

抄録

The purpose of this study is to make autonomous mobile robots act along human flow in a crowded environment such as stations and downtown areas. As a pre-stage, we propose a method to detect and tracking pedestrians using combining LiDAR point cloud and camera image. In the proposed method, pedestrians are first recognized using Darknet YOLO v2 from the camera image. Then projecting the point cloud data acquired by LiDAR on the camera image, extracting the point cloud of pedestrians only. Further, it is converted into an image overlooking the acquired point cloud, it performs tracking using a particle filter. In this paper, we describe the algorithm of the proposed method and results of implementation by autonomous mobile robot.