An Improved YOLO-V8 Algorithm for Rail Surface Defect Detection

Transportation by train is a backbone of countrywide infrastructure, and therefore, the issue of railway safety cannot be overlooked, Cracks, worn-out, and other imperfections on the outer surfaces of the rail tracks are some of the important reasons for the occurrence of accidents and train derailments. Manual analysis techniques are highly time- consuming and tend to have errors, and hence, are not appropriate for application at large railway networks. To address the issues, this project proposes a rail surface defect automated detecting system employing the YOLOv8 algorithm. A camera is fixed to a rail inspection prototype to generate continuous photographs of the rail track surface. The photographs are then analysed through computer vision and deep learning algorithms to automatically locate rail defects in real-time processing. The optimized YOLOv8 algorithm has high accuracy in rail defect detection with minimal processing time. Experimental results prove that the system is working well under changing lighting conditions and environmental settings. The system is effective in preventing manual rail inspection and is ideal in intelligent rail monitoring systems due to its contribution to rail safety.