DL-BIM: A Hybrid Deep Learning-Born Iterative Method for Electromagnetic Inverse Scattering Problems
IEEE Transactions on Microwave Theory and Techniques · 2026-08 · 74 (8) · 7163-7172
DOI: 10.1109/tmtt.2026.3703857 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Microwave Theory and Techniques, 2026-08. Verfasst von Yunyun Hu und 6 weiteren Autor:innen. Die Arbeit nennt 40 Literaturangaben. Gefördert von: Fundamental Research Funds for the Central Universities.
Abstract (Originalsprache)
This article proposes an unsupervised hybrid deep learning–Born iterative method (DL-BIM) for accurate electromagnetic inverse scattering imaging, in which a physics-driven U-Net (PD-UNet) is embedded into the Born iterative framework. The proposed method integrates the strengths of the Born iterative method (BIM) in addressing nonlinear inverse problems with the powerful capability of deep learning in solving optimization problems. It takes the rough image obtained by the first-order Born approximation (BA) as the network input, and employs the $L_{2}$ norm of the data equation as the network loss function to refine the image. To further improve the optimization performance, a modified U-Net architecture is introduced by incorporating a confined activation function, which restricts the search space within predefined bounds rather than an unbounded range, thereby improving the accuracy and stability of the ill-posed inverse problem. Extensive synthetic and experimental studies are conducted to examine the effectiveness of the DL-BIM. Compared with BIM and supervised deep learning approaches, the DL-BIM consistently demonstrates superior performance in reconstructing both the target geometry and permittivity distribution.
Förderung
- Fundamental Research Funds for the Central Universities
Zitieren
APA
Hu, Y., Wang, H., He, Q., Zhang, Y., Tong, M. S., Wang, D., & Sun, Q. (2026). DL-BIM: A Hybrid Deep Learning-Born Iterative Method for Electromagnetic Inverse Scattering Problems. IEEE Transactions on Microwave Theory and Techniques, 74(8), 7163-7172. https://doi.org/10.1109/tmtt.2026.3703857
Vancouver
Hu Y, Wang H, He Q, Zhang Y, Tong MS, Wang D, et al. DL-BIM: A Hybrid Deep Learning-Born Iterative Method for Electromagnetic Inverse Scattering Problems. IEEE Transactions on Microwave Theory and Techniques. 2026;74(8):7163-7172. doi:10.1109/tmtt.2026.3703857
BibTeX
@article{Hu2026,
author = {Hu, Yunyun and Wang, Hui and He, Qing and Zhang, Yuting and Tong, Mei Song and Wang, Dezhi and Sun, Qingtao},
title = {DL-BIM: A Hybrid Deep Learning-Born Iterative Method for Electromagnetic Inverse Scattering Problems},
journal = {IEEE Transactions on Microwave Theory and Techniques},
year = {2026},
volume = {74},
number = {8},
pages = {7163--7172},
doi = {10.1109/tmtt.2026.3703857}
}
EUVIMED