Automated Crack Detection on Train Tracks Using CNNs

Ensuring the safety and reliability of railways is vital for both operational efficiency and passenger safety. Traditional methods of rail inspection, though effective, often require significant time and human resources, and may not be feasible for large-scale or real-time monitoring. This paper introduces a solution that leverages deep learning, specifically the YOLOv9 (You Only Look Once version 9) model, to detect various defects in railway tracks. The approach uses a dataset of rail images with labeled defects, such as cracks, flakings, and missing bolts. Through careful preprocessing and data augmentation, the YOLOv9 model is trained to recognize these defects efficiently. Evaluation metrics, including precision, recall, and mean average precision (mAP), show promising results in defect detection accuracy. The model demonstrates the potential for real-time applications in automated rail maintenance systems, reducing the need for manual inspections. In future work, the system could be expanded to handle real-time video streams and even integrated with drone technology for enhanced monitoring capabilities. Keywords: Rail defect detection, YOLOv9, deep learning, automated inspection, computer vision.