Weld Defect Detection based on Improved Multi-Scale CNN with CBAM Attention

Authors:
Open Access
Publisher website
Google Scholar
Abstract
Welding is an important joint technology, but the weld defects have great influence on product quality. Because of the diversity of weld defect characteristics, weld defect detection is a complicated work in industry. However, it is a challenging task because weld defects can exhibit different shapes, sizes, positions, and contrasts in radiographic images. A weld defect identification method based on improved multi-scale CNN with CBAM attention (IMSACNN) is proposed. In IMSACNN, dilated Inception module is used to extract multi-scale convolutional feature graph and the convolutional block attention module (CBAM) is introduced into model on the space and channel, so as to strengthen the feature extraction ability. The experimental results on the image dataset of weld defects show that the proposed method is effective with the accuracy of 87.19%.
Funding Information
  • The National Natural Science Foundation of China (62172338,62072378)

This publication has 13 references indexed in Scilit: