EUVIMEDEUVIMED entdecken
Crossref

Self-Attention Mechanism-Enhanced Dictionary Learning for Sparse ISAR Imaging

Changyu Hu

IEEE Transactions on Aerospace and Electronic Systems · 2026 · 62 · 13604-13618

DOI: 10.1109/taes.2026.3710708 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Aerospace and Electronic Systems, 2026. Verfasst von Changyu Hu. Die Arbeit nennt 25 Literaturangaben. Gefördert von: General Program of Basic Science, Natural Science Foundation of the Jiangsu Higher Education Institutions of China, Light of Taihu” Key Technologies R&D Program of Innovation and Entrepreneurship Foundation.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Inverse synthetic aperture radar (ISAR) can provide the high-resolution images of noncooperative targets under all-weather day-and-night conditions. Dictionary learning (DL)-based sparse ISAR imaging methods have shown impressive imaging performance by exploiting the sparse representations of targets relative to their scenes. However, the imaging results of these methods may still exhibit incomplete target contours, especially under low-compression-ratio conditions. One essential reason lies in their limited capability to capture the global structural features of targets. To address this issue, this article first presents a self-attention mechanism (SAM)-enhanced DL model, termed SAM-DL, in which the SAM is employed to guide dictionary atom updates, enabling the dictionary to learn richer global structural features under low-compression-ratio conditions. Furthermore, an SAM-DL-based sparse ISAR imaging method is proposed, where a generalized Gaussian distribution assumption and a Lagrangian dual-ascent strategy are introduced to simplify the imaging process and improve convergence stability. Experimental results on both simulated and real ISAR datasets demonstrate that the proposed imaging method achieves more accurate reconstructions with clearer target contours and fewer artifacts than existing DL-based sparse ISAR imaging methods, which validates its effectiveness in maintaining the target’s global structural features under low-compression-ratio conditions.

Förderung

  • General Program of Basic Science
  • Natural Science Foundation of the Jiangsu Higher Education Institutions of China
  • Light of Taihu” Key Technologies R&D Program of Innovation and Entrepreneurship Foundation
  • Key Laboratory of Radar Imaging and Microwave Photonics Technology
  • Nanjing University of Aeronautics and Astronautics
  • Ministry of Education
  • Wuxi University Research Start up Fund for Introduced Talents

Zitieren

APA

Hu, C. (2026). Self-Attention Mechanism-Enhanced Dictionary Learning for Sparse ISAR Imaging. IEEE Transactions on Aerospace and Electronic Systems, 62, 13604-13618. https://doi.org/10.1109/taes.2026.3710708

Vancouver

Hu C. Self-Attention Mechanism-Enhanced Dictionary Learning for Sparse ISAR Imaging. IEEE Transactions on Aerospace and Electronic Systems. 2026;62:13604-13618. doi:10.1109/taes.2026.3710708

BibTeX
@article{Hu2026,
  author = {Hu, Changyu},
  title = {Self-Attention Mechanism-Enhanced Dictionary Learning for Sparse ISAR Imaging},
  journal = {IEEE Transactions on Aerospace and Electronic Systems},
  year = {2026},
  volume = {62},
  pages = {13604--13618},
  doi = {10.1109/taes.2026.3710708}
}
Kostenlos starten →