A Detection Algorithm for Metal-Bearing Roller Microcracks with Global Contrast and Threshold Region Growth

To address the problem of incomplete extraction for microcracks due to low contrast and blurred edges in metal-bearing roller line microcrack images, a detection algorithm for low-contrast line microcracks based on global contrast enhancement and threshold region growth segmentation is proposed. A global contrast enhancement algorithm is constructed by analyzing low-contrast characteristics in metal-bearing roller line microcrack images. Defining a multi-scale Gaussian function for contrast stretching and adjusting contrast parameters to complete contrast enhancement, combined with the metal-bearing roller microcrack edge blurring problem, the seeds are selected according to the gradient criterion. Meanwhile, the region growth of the gray value criterion is designed to complete the extraction for line microcracks. The results show that the metal-bearing roller line microcrack images are enhanced by global contrast. The enhanced images achieve an SSIM value of 0.9590 and a mean error of 0.008. The extraction precision of the metal-bearing roller line microcrack is 98.503%.