Railway Obstacle Intrusion Warning Mechanism Integrating YOLO-Based Detection and Risk Assessment

Railroad obstacle intrusion can significantly threaten rail safety operations, especially with rapidly developed high-speed railways. Object detection has played an important role in rail safety operations. Nowadays, detection technologies have been employed to detect accurately and in real-time with a lot of effort from scholars. However, these detection methods still suffer from the inability to distinguish dangerous obstacles and deliver timely warnings based on their severity. To address these issues, this study innovatively proposes an early warning approach that fuses the detection of events and information on foreign objects’ severities. Specifically, the developed method can detect the objects with the state-of-the-art YOLOv5 model and extract the track area by the improved-railway-detection method. The dangerous objects are distinguished and the warning level is determined according to the obstacle's position and severity. The experimental results demonstrate the feasibility and practicability of the proposed method and its feasibility in relatively low-light environments. The integration of more classifications for risk assessment in the future will contribute to a more accurate and reliable warning basis for decision-makers.