Research on Machine Vision Optical Component Surface Defect Anomaly Detection System

This paper discusses using machine vision technology to identify defects on the surface of optical components and extract effective features from them. Firstly, geometric feature parameters of the defect images are analyzed and extracted based on the characteristics of various surface defects on optical components. Secondly, the attributes and differences between Principal Component Analysis (PCA) and Independent Component Analysis (ICA) are analyzed, and PCA is selected to reduce the dimensionality of the extracted feature parameters, removing redundant feature information, reducing the size of classifier input parameters, and improving the classifier’s classification efficiency. The results show that, by selecting distinctive feature parameters based on the characteristics of the objects to be classified at each level and considering the limited number of samples, the average classification accuracy is 92.2%. Scratches are easily distinguished from other surface defects, while identification rates for spots and dust are relatively lower.