Identification of welding defects by combining magnetic-optical imaging and infrared imaging

Focusing on spot welding defects, a method for identifying welding defects by combining magneto-optical imaging (MOI) and infrared imaging (IRI) is investigated based on the thermoelectric effect and the Faraday magneto-optical (MO) rotation effect. A detection platform is constructed to collect magneto-optical and infrared (IR) images of defect-free samples as well as common and more severe defects such as cracks, pits, and incomplete fusion. The method of enhancing MO and IR images is employed by utilizing fast non-local means filtering, image normalization, and image sharpening techniques. Adaptive pixel weighted fusion is applied to combine the MO and IR images, yielding fused images. Subsequently, basic probability assignments for each class and uncertainties for each modality are obtained through the linked Dirichlet distribution. Finally, Dempster’s combination rule is employed for decision fusion, enabling the classification and identification of welding defects. Experimental results show that the proposed method effectively exploits the advantages of MOI and IRI, thereby improving the accuracy of welding defect identification.