Instance Segmentation Based Non-Destructive Inspection of High-voltage Cable Defects

The defect inspection of high-voltage cable is an important guarantee for the safe operation of cable. Traditional manual inspection is complex and costly. To solve this problem, this research combine X-ray NDT(non-destructive testing) with deep learning to propose an edge refinement mask network to improve the efficiency and accuracy of cable defect inspection. Specifically, in order to improve the accuracy of network inspection, this research employ a variety of data enhancement methods to extend the X-ray image dataset. For the cable defect edge, this study add the edge refinement operation after the basic example segmentation model, so that the segmented defect edge is more realistic.The experimental results show that the network defect detection accuracy proposed in this study reaches 91.6% and the detection speed reaches 54.7fps. Compared with the general defect inspection methods, the method in this paper has a greater improvement in accuracy and convenience.