Research on surface scratch detection methods for mechanical parts in industrial quality inspection

Scratch detection on metal surfaces holds significant importance. Conventional manual inspection methods suffer from low efficiency and high subjectivity, necessitating automated image detection solutions. A scratch detection method based on the OpenCV platform is proposed, employing the Canny edge detection operator with adaptive thresholding. Initially, images undergo preprocessing through greyscale conversion. Subsequently, the CLAHE algorithm enhances scratch features, followed by Gaussian filtering to eliminate image noise. Finally, the Canny operator with Otsu's adaptive thresholding extracts scratch edges. The processed image is then subjected to dilation, with scratch regions annotated using blue bounding boxes. Experimental results demonstrate the method's high detection accuracy and practical applicability.