Railway Track Surface Defect Detection: Advancements in Computer Vision and Image Processing

The traditional method of detecting defects on track surfaces relies on manual inspection, which is subjective, inefficient, and time-consuming. However, in recent years, computer vision and image processing technology have been increasingly utilized in railway defect detection. This has significantly improved the accuracy and efficiency of detection.

Bai et al. combined an enhanced Faster R-CNN network with support vector machines to detect defects in railroad fasteners. Qi et al. developed an improved lightweight detection network specifically for detecting defects in railroad fasteners. Liu et al. proposed a cascade detection method based on a depth neural network (DCNN) and an enhanced Fast RCNN detection method. This approach not only enables fastener detection in real-time but is also suitable for detecting defects in unbalanced samples.

Ni et al. introduced a visual inspection system (VIS) comprising an image acquisition system and a rail surface inspection system. This system enables the detection of indicated defects in both straight and curved rails. Nieniewski et al. integrated morphological operation methods into a neural network for rail defect detection and shape extraction. Wei et al. put forward a multi-objective defect detection network called TLMDDNet, which allows for the simultaneous detection of rail fasteners and rail surface defects.