Study on Detection of Foreign Objects in Rail Transport Sections Based on Radar Points Clouding and Image Recognition Technology

Addressing issues like low detection rates, prolonged response times, and high false alarm rates in detecting foreign objects that fall within rail transit sections under complex weather conditions, this paper proposes an algorithm for foreign object detection that combines radar and visual analysis. First, based on the laser radar and the intelligent PTZ camera model, the point cloud data and image data are established to acquire high-precision spatial information, including the coordinates, shape features, and size of the orbital foreign body. Second, this paper proposes a multi-frame point cloud superposition strategy to eliminate the limitation of single-image detection technology for open-set object recognition. Third, according to the high-precision characteristics of lidar, a radar-vision fusion model considering the influence of video image analysis technology and IoT sensing technology is constructed to describe the foreign object invasion more accurately. Finally, in this study, a foreign object detection system for rail transport sections based on a radar-vision fusion algorithm is built. Simulation and experimental review demonstrate the effectiveness of the radar vision fusion algorithm proposed in this paper, achieving the goal of an average detection rate of foreign objects with a size of 10 cm * 10 cm within 100 meters, exceeding 80%, and alarm time of less than 5 seconds.