EUVIMEDEUVIMED entdecken
Crossref

PGDNet: Prior Guided Dual-Domain Fusion Network for Infrared Small Target Detection

Zibo Pang, Hanlin Qin, Xupei Zhang, Juzheng Liu

IEEE Transactions on Circuits and Systems for Video Technology · 2026-11 · 36 (11) · 17761-17775

DOI: 10.1109/tcsvt.2026.3719264 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Circuits and Systems for Video Technology, 2026-11. Verfasst von Zibo Pang und 3 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 55 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Fundamental Research Funds for the Central Universities, Nature Science Basic Research Program of Shaanxi.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Infrared small target detection remains unreliable in cluttered scenes. Small target structures are easily degraded during cross-scale feature propagation, and their weak responses are often overwhelmed by background-dominant spectra. In addition, attention mechanisms tend to be biased toward salient clutter when target saliency is extremely low. We propose a Prior Guided Dual-Domain Fusion Network for infrared small target detection, in which semantic priors from a vision foundation model guide adaptive spatial–spectral fusion for robust target localization in cluttered scenes. PGDNet designs a Dynamic Structure-Constrained Diffusion Mechanism (DSCDM) to stabilize cross-scale propagation and preserve the spatial integrity of small targets. Built upon this structure-consistent representation, PGDNet develops Spectral Channel Selection Attention (SCSA) to enhance target-relevant frequency responses and suppress background interference. On this basis, cross-region attention priors are learned by fine-tuning DINOv3 and are integrated through a curriculum adaptive fusion(CAF) strategy to regulate prior influence and maintain stable guidance during training. Extensive experiments on three public datasets demonstrate that PGDNet consistently achieves the best mIoU (78.83 / 94.64 / 75.50 in NUAA-SIRST, NUDT-SIRST, and IRSTD-1K), while maintaining a favorable balance between $P_{d}$ and $F_{a}$ under complex backgrounds. Our code will be made publicly available at https://github.com/Eureka10086/PGDNet

Förderung

  • National Natural Science Foundation of China
  • Fundamental Research Funds for the Central Universities
  • Nature Science Basic Research Program of Shaanxi
  • Ningbo Natural Science Foundation
  • Ministry of Education Joint Foundation
  • Shaanxi Province Science and Technology Plan Project
  • Xi’an City Science and Technology Plan Project

Zitieren

APA

Pang, Z., Qin, H., Zhang, X., & Liu, J. (2026). PGDNet: Prior Guided Dual-Domain Fusion Network for Infrared Small Target Detection. IEEE Transactions on Circuits and Systems for Video Technology, 36(11), 17761-17775. https://doi.org/10.1109/tcsvt.2026.3719264

Vancouver

Pang Z, Qin H, Zhang X, Liu J. PGDNet: Prior Guided Dual-Domain Fusion Network for Infrared Small Target Detection. IEEE Transactions on Circuits and Systems for Video Technology. 2026;36(11):17761-17775. doi:10.1109/tcsvt.2026.3719264

BibTeX
@article{Pang2026,
  author = {Pang, Zibo and Qin, Hanlin and Zhang, Xupei and Liu, Juzheng},
  title = {PGDNet: Prior Guided Dual-Domain Fusion Network for Infrared Small Target Detection},
  journal = {IEEE Transactions on Circuits and Systems for Video Technology},
  year = {2026},
  volume = {36},
  number = {11},
  pages = {17761--17775},
  doi = {10.1109/tcsvt.2026.3719264}
}
Kostenlos starten →