Computer Vision Based Crack Detection of Railway Track

In India, trains are vital mode of transportation for millions, the safety of railway infrastructure is of utmost importance. In this work, a detection system has been developed using the YOLOv5 algorithm to overcome the challenges of traditional railway track inspection methods, which are manual, labor-intensive, inconsistent, and often hazardous for workers. This system enhances safety and efficiency by employing a web camera mounted on a self-propelled vehicle to capture continuous images of railway tracks. These images are pre-processed and then analyzed in real-time using a YOLOv5-trained model, which excels in accurately detecting cracks. Ultrasonic sensor is used to detect obstacles and avoid collisions. Upon detecting a crack and unwanted objects on the track, the system takes immediate action by halting the vehicle ensuring precise and timely responses to potential issues. Simultaneously, an alert message is sent to authorize personnel. The system achieves precision of 94.74%, recall with 90% and accuracy with 92.5%.