In industrial pipeline systems, quality monitoring of steel pipes and welds is a critical component to ensure safe operation. Utilizing deep learning techniques to analyze X-ray images can efficiently identify weld defects, such as porosity, cracks, and inclusions, significantly enhancing the accuracy and efficiency of non-destructive testing. In response to the existing weld defect detection models' insufficient feature extraction and lack of handling diversity, this paper introduces a weld defect identification method based on the Hierarchical Attention Fusion Network (HAFNet). Initially, a Dilated Hierarchical Attention Mechanism (DHAM) is employed to capture multi-scale global and local information, thereby enhancing the focus on key features of different scales and effectively addressing the issue of large intra-class variability and small inter-class differences in defects. Subsequently, a Residual Fusion Module (RFM) is introduced, which adaptively learns the feature weights of different encoding layers and fully utilizes contextual information during the decoding phase to suppress the complex background interference of weld images. Finally, through a Multi-Level Feature Fusion Module (MFFM), the decoded network's multi-layer features are strengthened by a fusion mechanism, enhancing the interaction and complementarity between different levels of features, reducing the model's sensitivity to noise and non-critical information, and further enhancing the model's recognition accuracy and robustness.

Ruixiang Li, Shanwen Zhang, Lei Huang, Mingda Yang, Chengyu Hu. Enhanced Image Segmentation-Based Detection Technique for X-ray Film Images of Weld Seams. Academic Journal of Computing & Information Science (2024), Vol. 7, Issue 4: 118-128. https://doi.org/10.25236/AJCIS.2024.070416.

[1] Lin Z, Yingjie Z, Bochao D, et al. Welding defect detection based on local image enhancement[J]. IET Image Processing, 2019, 13(13):2647-2658. 

[2] Xu Z, Wu M, Fan W. Sparse-based defect detection of weld feature guided waves with a fusion of shear wave characteristics [J]. Measurement, 2021, 174:109018.

[3] Abdelkader R, Ramou N, Khorchef M, et al. Segmentation of x-ray image for welding defects detection using an improved Chan-Vese model[J]. Materials Today: Proceedings, 2021, 42:2963-2967.

[4] Yan Z H, Xu H, Huang P F. Multi-scale multi-intensity defect detection in ray image of weld bead [J]. NDT & E International, 2020, 116:102342.

[5] Malarvel M, Singh H. An autonomous technique for weld defects detection and classification using multi-class support vector machine in X-radiography image [J]. Optik, 2021, 231:166342.

[6] Wang H, Li J, Liu L. Process optimization and weld forming control based on GA-BP algorithm for riveting-welding hybrid bonding between magnesium and CFRP [J]. Journal of Manufacturing Processes, 2021, 70:97-107.

[7] Park J-K, An W-H, Kang D-J. Convolutional Neural Network Based Surface Inspection System for Non-patterned Welding Defects [J]. International Journal of Precision Engineering and Manufacturing, 2019, 20(3):363-374.

[8] Miao R, Gao Y, Ge L, et al. Online defect recognition of narrow overlap weld based on two-stage recognition model combining continuous wavelet transform and convolutional neural network[J]. Computers in Industry, 2019, 112:103115.

[9] Xiao M, Yang B, Wang S, et al. A feature fusion enhanced multiscale CNN with attention mechanism for spot-welding surface appearance recognition [J]. Computers in Industry, 2022, 135:103583.

[10] Chen Z, Huang G, Lu C, et al. Automatic recognition of weld defects in ToFD D-scan images based on faster R-CNN[J]. Journal of Testing and Evaluation, 2020, 48(2):811-824.

[11] Zhang K, Shen H. Solder joint defect detection in the connectors using improved faster-rcnn algorithm [J]. Applied Sciences, 2021, 11(2):576.

[12] Jiang D, Li G, Tan C, et al. Semantic segmentation for multiscale target based on object recognition using the improved Faster-RCNN model[J]. Future Generation Computer Systems, 2021, 123:94-104.