Magneto-Optical Imaging Feature Extraction of Angle Welding Defects Based on Polarization Information

In order to improve the effective data quantity of corner weld defect magneto-optical imaging feature extraction, a method based on polarization information is proposed. The magneto-optical imaging principle of corner welding defects is analyzed, and the magneto-optical image of defects is collected. Based on this, the MOO image is preprocessed by smoothing filter, image enhancement, background removal and specular highlights respectively, and the result of feature extraction is obtained. Compared with the traditional feature extraction method, the effective data volume of the design feature extraction method is improved by 19.535MB.