A Point Cloud Gridding Method for Detecting Rail Intruding Foreign Object

Intruding foreign objects (IFO) pose a serious threat to railway safety operations. Existing detection algorithms often suffer from significant false alarms and missed detections, rendering them ineffective in practical IFO detection. To address this issue, we propose a point cloud gridding method for detecting rail IFO. Firstly, this method provides effective IFO detection hardware by selecting an appropriate lidar angular resolution based on the requirements for detecting IFO. Secondly, we map pillar point clouds to Bird's Eye View (BEV) to form detection grids, enabling effective IFO detection and improving detection accuracy and recall rates. Extensive experiments and real-world applications have demonstrated the effectiveness of this method.