EUVIMEDEUVIMED entdecken
Crossref

False Alarm Rate-Controllable Maritime Target Detection via Cross-Attention Transformer Network

Qingsong Jiang, Wenjing Zhao, Weiqing Zhao, Guolong Cui

IEEE Sensors Letters · 2026-07 · 10 (7) · 6006304-6006304

DOI: 10.1109/lsens.2026.3696409 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Sensors Letters, 2026-07. Verfasst von Qingsong Jiang und 3 weiteren Autor:innen. Die Arbeit nennt 19 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

To address degraded maritime target detection and false alarm rate (FAR) control under strong sea clutter, this letter proposes FARC-CAT, a FAR-controllable detector based on a dual-encoder cross-attention Transformer. An energy-frequency dual-branch encoder is designed for feature extraction, followed by bidirectional cross-attention and gated fusion to enhance feature interaction. In addition, a hybrid loss with a learnable threshold is introduced to constrain FAR during training. Experiments on publicly available measured datasets show that FARC-CAT outperforms representative methods in low signal-to-clutter ratio scenarios while maintaining a low FAR. In particular, under a preset FAR of $10^{-4}$, it achieves an average detection probability of 91.97%, outperforming the existing representative methods.

Zitieren

APA

Jiang, Q., Zhao, W., Zhao, W., & Cui, G. (2026). False Alarm Rate-Controllable Maritime Target Detection via Cross-Attention Transformer Network. IEEE Sensors Letters, 10(7), 6006304-6006304. https://doi.org/10.1109/lsens.2026.3696409

Vancouver

Jiang Q, Zhao W, Zhao W, Cui G. False Alarm Rate-Controllable Maritime Target Detection via Cross-Attention Transformer Network. IEEE Sensors Letters. 2026;10(7):6006304-6006304. doi:10.1109/lsens.2026.3696409

BibTeX
@article{Jiang2026,
  author = {Jiang, Qingsong and Zhao, Wenjing and Zhao, Weiqing and Cui, Guolong},
  title = {False Alarm Rate-Controllable Maritime Target Detection via Cross-Attention Transformer Network},
  journal = {IEEE Sensors Letters},
  year = {2026},
  volume = {10},
  number = {7},
  pages = {6006304--6006304},
  doi = {10.1109/lsens.2026.3696409}
}
Kostenlos starten →