Abstract

In the manufacturing process of razor blades, the presence of surface defects directly impacts their quality and lifespan. Manual detection of these defects is not only inefficient but also susceptible to external interference. To achieve real-time and stable detection of blade defects, a method based on video analysis was proposed in this paper. Raw surface images of the blades were captured using a computer vision system, and binarization was performed using Otsu's algorithm. Subsequently, contours were identified and preliminarily analyzed using the CCL algorithm to obtain Regions of Interest (ROI). The ROI was then sorted, and subtle defects were filtered out through morphological analysis. A mask was obtained by comparing the ROI with a background model constructed based on frame averaging. Through mask analysis, defect types were determined, achieving efficient detection of surface defects on razor blades. Experimental results demonstrated that the proposed method enabled real-time detection of multiple defect types simultaneously.