Rail transport, being a global mode of transportation, effectively facilitates long-distance travel for passengers worldwide, benefiting travelers from all corners of the globe. However, derailment poses a significant challenge to the industry’s safety and reliability. Regular inspection of tracks help to identify any signs of wear, damage, or defects that could lead to derailments. These inspections involve examining track components such as rails, joints, switches, and fastenings, for any abnormalities or potential hazards. Inadequate ightening or loosening of nuts and bolts can occur due to vibrations, temperature variations, or inadequate maintenance practices. Loose or missing fasteners can cause track misalignment and instability, potentially leading to derailments.
The main objective of this work is to detect various components of joint bars including nuts, bolt heads, and identifying specific cases such as missing nuts and holes. This study examines the latest advancements in computer vision integrated into You Only Look Once version 5 (YOLOv5), a state-of-the-art object detection framework. To achieve this, a model based on YOLOv5 is constructed using input images captured by drone, which provide high-resolution images from a bird’s eye view. The use of drones offers significant advantages over traditional inspection methods such as improved safety, cost-effectiveness, and the ability to cover large areas quickly. The model is then trained using the captured images, and its accuracy is subsequently evaluated. Once the rail components were identified and consolidated, further data integration and analysis took place, resulting in a significant enhancement of the overall accuracy to 97 percent. The results show how good the system is at recognizing problems and their components accurately.