Automatic Extraction Method of Weld Weak Defect Features for Ultra-High Voltage Equipment

To solve the problems of low precision weak feature extraction, heavy reliance on labor and efficiency extraction in X-ray weld detection image ultra-high voltage (UHV) equipment key parts, an automatic algorithm is proposed. Firstly, original denoised while retaining characteristic information defects by proposed monostable stochastic resonance method. Then, binarization achieved combining Laplacian edge Otsu threshold segmentation. Finally, identification defect area realized based sequential traversal binary tree. Several analysis dimensions are established for UHV including area, perimeter, slenderness ratio, duty cycle, etc. The experiment using actual production site shows that method can effectively extract further provide reference decision-making.