关键词: 钢轨表面缺陷, 灰度标准差, 多尺度灰度对比图, 鲁棒性

Abstract: When machine vision technology was used to detect the surface defects of rails, the background lighting was complex, the relative position of the on-board detection equipment and the rail changes, etc., which seriously affected the defect detection accuracy. Therefore, a rail surface area localization algorithm was proposed based on gray standard deviation and projection integration, and an enhancement algorithm was also presented based on multi-scale gray-scale contrasts. The gray standard deviation was used to eliminate the interferences of complex backgrounds, and the precise positions of the orbital surfaces was obtained through the projection integral, then, the gray-scale contrasts in different scale spaces were obtained to convert the rail surface areas into a gray contrast maps for defect enhancements, finally, iterative threshold segmentation method was used to extract rail surface defects. Experiments show that, under several different shooting conditions, the missing detection rate of the rail surface defect detection algorithm proposed is less than 6%， and the accuracy rate is more than 93%. This algorithm has high robustness for the detection of surface defects on seamless rails in high-speed ballast tracks.

Key words: rail surface defect, gray standard deviation, multi-scale grayscale contrast diagram, robustness