Two-dimensional Gabor filter-based  belt tear detection

As the existing belt tear detection method based on machine vision technology being used for images with complex background texture, it is easy to misjudge the defect areas with inconspicuous tear marks relative to the background texture as non-defects, and the detection results are noisy and difficult to identify. In order to solve the above problems, a belt tear detection method based on two-dimensional Gabor filter  is proposed. The method uses Gabor filtering to process the belt images and obtain multiple Gabor filtered processing images. Through the Gabor optimization selection method, a new cost function is constructed based on the coefficient of variation, and the optimal filter channel is selected to highlight the texture characteristics of the tear area. The Sobel operator is used to extract the texture characteristics of the tear area in the horizontal and vertical directions respectively so as to obtain the gradient images of the two directions. The self-propagation normalization operation is applied on the obtained gradient images to enhance the texture information, and the pixel-weighted average method is used to fuse the two images. The obtained fused images are segmented by adaptive threshold binarization method, and morphological technology is used to perform belt tear detection on the images to be detected. The detection results show that the improved Gabor optimization selection method has a lower miss detection rate than the Gabor optimization selection method and the Sobel operator-based longitudinal tear detection method. The method can detect all defects in the belt defect images with complex background texture. The detection results are clear and the contour characteristics of the tear area are preserved relatively well.