Research on X-ray welding image defect detection based on convolution neural network

Abstract

In order to improve the efficiency of X-ray welding image defect recognition, it is proposed to use the deep learning network to identify welding defects. Based on the analysis of X-ray weld defect image characteristics, the convolutional neural network template and the number of layers are determined. By constructing a deep learning network structure that simulates the principle of visual perception, the steps of feature extraction of weld defect images are avoided. Also the deep learning network can directly determine whether the suspected defect image is a linear defect, circular defect or noise. The designed system can automatically learn the complex depth features in the X-ray weld defect image. The actual calculation shows that the proposed method is feasible and effective.