Enhancing Corrugation Detection via GAN-Based Data Augmentation

With the rapid development of high-speed rail and rail transportation, the health of the rails is critical to operational safety. However, conventional manual inspection methods are inefficient and susceptible to subjective influences, which cannot meet the needs of modern railway systems. In this study, the generative adversarial network (GAN) technology was combined to generate rail corrugated wear images to solve the problem of insufficient data samples, and the improved you only look once (YOLO) algorithm was used for automatic defect detection. This article compares the images generated by multiple GANs (such as WGAN, WGAN-GP) and conducts experimental analysis of the detection effects of different versions of the YOLO model. The results show that the proposed method not only improves the diversity of training data but also significantly improves the accuracy and efficiency of rail defect detection. In this study, an efficient and reliable rail defect detection scheme is proposed, which provides technical support for the operation safety of rail transit.