Two-stage detection algorithm of train key bolts based on YOLO and Faster R-CNN

With the development of economy, subway trains have become the first choice for most people to travel. Bolts play an important role in connecting subway trains, so rapid detection of bolt loosening and falling off is a very forced problem to be solved. In the field of computer vision, small target recognition has always been considered one of the most challenging problems. Based on Faster R-CNN and YOLOv8, a two-stage algorithm named JY-YOLO algorithm is proposed in this paper. In the first phase of the algorithm, Faster R-CNN has used the input images of any size of subway to locate and segment the key areas where the bolts are located -Axle Box Cover and Shock Absorber. The segmented image will be fed into the second stage neural network FasterYOLO ResBlock_CBAM in different channels to accurately locate the location of the bolt and detect whether it falls off or becomes loose. The accuracy of the algorithm for each item is more than 98.9%, and it can accurately locate the bolt fault.