Defect detection of tensile clamp press-fitting in power transmission lines based on X-ray image preprocessing and deep networks

Aiming at the problem of small size of crimped part and background noise in the X-ray image of tensioning clamps, a crimping defect detection method for transmission line tensioning clamps based on X-ray image preprocessing combined with deep network is proposed. Firstly, X-ray non-destructive testing technology is used to obtain tension clamp inspection image data containing crimping defects, secondly, the image data is preprocessed using 3D block filtering algorithm and homomorphic filtering algorithm to remove the background noise and restore the image detail features, and then the preprocessed image is input into the joint detection network based on CenterNet and YOLOv7, and the crimping defects are localized in CenterNet network and intercepted in the crimped part. The CenterNet network locates the crimped part, intercepts the image of the crimped part and inputs it into the YOLOv7 network to complete the detection of crimped defects. The experimental results show that the accuracy of the proposed method reaches up to 92.40%, and the average value of AP reaches 91.38%, which has good detection effect.