Nondestructive Testing of Wire Ropes Based on Image Fusion of Leakage Flux and Visible Light

Magnetic flux leakage (MFL) detection has the advantages of obvious target defect and accurate positioning. The MFL information on the surface of the wire rope can be converted into magnetic images so that the axial and circumferential position of the wire rope defects can be displayed more intuitively. The visible image has rich texture information. By fusing the visible and magnetic images, we can make full use of their information and the defect recognition rate can be improved. In this paper, the original magnetic leakage data are first denoised by wavelet soft threshold, and the corresponding visible image is processed by homomorphic filtering to eliminate the interference of light. Then, the magnetic image and visible image are fused at feature level. The features of magnetic image and visible image are extracted and fused, and then principal component analysis is carried out to reduce dimensions. The fusion feature vector is input into back-propagation network for recognition and is compared with the magnetic image features alone. Experimental results show that when the error is allowed to be 0.9%, the defect recognition rate after image fusion is 5.27% higher than the magnetic image.