A Detection Algorithm of Rail Surface Defects Based on Deep Learning

In the field of railway transportation, the integrity of the rails is a core element in ensuring the safe operation of trains. As the railway network continues to expand and the transportation load increases, the detection of surface defects on the rails becomes particularly critical. Neglecting defects such as cracks, wear, and corrosion could lead to catastrophic consequences. This study is dedicated to developing an advanced rail surface defect detection algorithm based on artificial intelligence, significantly enhancing the accuracy and efficiency of detection. We propose an innovative “Enhanced Feature Fusion U-Net” (EFF-U-Net) model, which, through its designed skip connections, achieves efficient fusion of features, thereby greatly improving the recognition ability of the model for details on the rail surface. Faced with the scarcity of rail surface defect datasets, we have employed data augmentation techniques to effectively expand the sample size and enhance the model's generalization capabilities. Through experimental validation on the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{RSDDS}+\text{dataset}$</tex>, our method has demonstrated exceptional performance, significantly improving the accuracy and identification capability of defect detection.