Obstacle Detection Method using On-Train Forward-facing Cameras and Sensors

Abstract

Developing sensing technology that can reliably detect obstacles on the track well ahead of a train is essential for advancing on-train forward-facing track monitoring. Therefore, we developed a method for detecting obstacles in front of trains using cameras and sensors. In our method, obstacles such as people and vehicles are detected by combining multiple sensors within a detection area around the tracks. In this paper, we report on the results of a study of sensor configurations suitable for obstacle detection, the detail of our detection algorithm using cameras and LiDARs or a stereo camera, and the results of detection performance evaluation according to distance between train cab and object.