Research on Active Obstacle Detection Algorithm of Rail Train Based on Multi-sensor

Aiming at the problems of false detection, missed detection and poor environmental adaptability when a single sensor detects obstacles in front of the rail train, an active obstacle detection algorithm for rail transit based on multi-sensor is proposed. Firstly, the track segmentation point set extracted by machine vision algorithm is used to fit the track by least square method to construct the driving envelope model. Secondly, the point cloud based on CFSFDP clustering is projected onto the image plane through the calibration of lidar and camera. The projected lidar clustering point cloud and camera image are combined with the driving envelope model to detect obstacles twice. Finally, based on the D-S evidence theory, the decision of the two detection results is made to overcome the detection limitations of a single sensor and improve the detection accuracy. The experimental results show that the algorithm has higher detection accuracy than the single visual or lidar sensor detection algorithm, and can effectively detect the sudden intrusion obstacles within the envelope boundary in front of the train.