https://doi.org/10.1140/epjp/s13360-025-06423-9
Regular Article
Swin-DCIR based for transmission image reconstruction of TGS nuclear waste drums
1
School of Computer Science and Engineering, Sichuan University of Science and Engineering, 644005, Yibin, China
2
School of Physics and Electrical Engineering, Sichuan University of Science and Engineering, 644005, Yibin, China
3
School of Nuclear Technology and Automation Engineering, Chengdu University of Technology, 610059, Chengdu, China
Received:
4
January
2025
Accepted:
13
May
2025
Published online:
3
June
2025
In the process of nuclear waste disposal and nuclear facility decommissioning, the amount of low-and-intermediate level radioactive waste will increase significantly. If it is not handled properly, it will threaten the development of the nuclear industry and human safety. Tomographic gamma scanning (TGS) technology is an effective method for non-destructive detection of low-and-intermediate level radioactive waste in sealed spaces, of which transmission imaging is an important part of TGS. The TGS transmission image reconstruction methods based on algebraic iterative algorithms and residual neural networks have made significant progress in improving image clarity and reconstruction speed, but they still face the problems of blurred details at the image edges, large noise bias, and poor objective evaluation indexes. Moreover, the quality of transmission image reconstruction directly impacts the reconstruction quality of subsequent emission images. Therefore, it becomes especially important to investigate a method that can preserve edge details, reduce noise, and improve reconstruction quality more effectively. In this paper, we propose an algorithmic model Swin-DCIR (Image Restoration Network with Swin Transformer and Dynamic Cross-Attention) for TGS transmission image reconstruction. The model introduces linear variable kernel convolution module and dynamic cross-attention network module in the shallow feature extraction module and deep feature extraction module, respectively. The introduced modules enhance the model's attention to global features and local details and improve the ability to capture details. In order to verify its significant advantages in TGS transmission image reconstruction, we conducted evaluation experiments on a homemade TGS transmission image dataset, and the experimental results show that the Swin-DCIR model has higher reconstruction accuracy, lower noise, and fewer parametric quantities in the dataset task, and the reconstructed image is visually clearer in terms of both the edge structure and the overall contour.
Copyright comment Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
© The Author(s), under exclusive licence to Società Italiana di Fisica and Springer-Verlag GmbH Germany, part of Springer Nature 2025
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
