Rail Surface Defect Detection Based on Image Enhancement and Improved YOLOX

During the long and high-intensity railway use, all kinds of defects emerge, which often produce light to moderate damage on surface, adversely affects stable operation trains even endangers safety travel. Currently, models for detecting rail surface are ineffective, self-collected images have poor illumination insufficient defect data. In aforementioned problems, this article suggests an improved YOLOX image enhancement method defects. First, a fusion algorithm is used in HSV space process steel rail, highlighting enhancing background contrast. Then, paper uses more efficient faster BiFPN feature neck structure YOLOX. addition, it introduces NAM attention mechanism increase expression capability. The experimental results show that detection using improves mAP network by 2.42%. computational volume increases, but speed can still reach 71.33 fps. conclusion, upgraded model detect flaws with accuracy speed, fulfilling demands real-time detection. lightweight deployment terminals also has some benefits.