EUVIMEDEUVIMED entdecken
Crossref

DCA-YOLO: A Dynamic Cross-Attention Network for Small Vehicle Detection in UAV Aerial Imagery

Chaojun Dong, Yizhi Wang, Yikui Zhai, Ye Li, Xiankun Liu, Chaoyun Mai, Jianhong Zhou, Pasquale Coscia, Angelo Genovese, C. L. Philip Chen

IEEE Transactions on Aerospace and Electronic Systems · 2026 · 62 · 15031-15047

DOI: 10.1109/taes.2026.3715596 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Aerospace and Electronic Systems, 2026. Verfasst von Chaojun Dong und 9 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 67 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Guangdong Province Key Disciplines Scientific Research Capacity Enhancement Project, Guangdong Higher Education Innovation and Strengthening School Project.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

To address challenges in uncrewed aerial vehicle (UAV) aerial imagery, such as complex backgrounds, drastic scale variations, and partial occlusions. We propose dynamic cross-attention YOLO (DCA-YOLO), a high-performance vehicle detection framework featuring a newly designed neck architecture named the DAC path aggregation network. Our approach introduces three synergistic modules in the neck to enhance feature expressiveness for small vehicle detection: 1) dynamic weighted adaptive feature fusion; 2) cross-layer dynamic interaction; and 3) coordinate-aware dynamic attention. These modules synergistically improve multiscale representation, seamlessly integrate shallow details with deep semantics, and dynamically emphasize key object features. To this end, we leverage the efficient and powerful FasterNet as the backbone to achieve an effective balance between computational overhead and detection accuracy. Experiments on the VisDrone and uncrewed aerial vehicle high scale (UAV-HS) datasets demonstrate that DCA-YOLO achieves an optimal accuracy-efficiency tradeoff among compact detectors. On the challenging VisDrone dataset, DCA-YOLO significantly outperforms the baseline YOLOv8n, boosting mAP$_{50}$ from 57.2% to 60.9%, mAP$_{50:95}$ from 38.5% to 41.6%, and mAP$_{s}$ from 14.3% to 16.1%.

Förderung

  • National Natural Science Foundation of China
  • Guangdong Province Key Disciplines Scientific Research Capacity Enhancement Project
  • Guangdong Higher Education Innovation and Strengthening School Project
  • Wuyi University Hong Kong and Macao Joint Research and Development Fund
  • Jiangmen Science and Technology Commissioner Research
  • WYU-BUAA Joint Laboratory of Aerospace Information Technology

Zitieren

APA

Dong, C., Wang, Y., Zhai, Y., Li, Y., Liu, X., Mai, C., Zhou, J., Coscia, P., Genovese, A., & Chen, C. L. P. (2026). DCA-YOLO: A Dynamic Cross-Attention Network for Small Vehicle Detection in UAV Aerial Imagery. IEEE Transactions on Aerospace and Electronic Systems, 62, 15031-15047. https://doi.org/10.1109/taes.2026.3715596

Vancouver

Dong C, Wang Y, Zhai Y, Li Y, Liu X, Mai C, et al. DCA-YOLO: A Dynamic Cross-Attention Network for Small Vehicle Detection in UAV Aerial Imagery. IEEE Transactions on Aerospace and Electronic Systems. 2026;62:15031-15047. doi:10.1109/taes.2026.3715596

BibTeX
@article{Dong2026,
  author = {Dong, Chaojun and Wang, Yizhi and Zhai, Yikui and Li, Ye and Liu, Xiankun and Mai, Chaoyun and Zhou, Jianhong and Coscia, Pasquale and Genovese, Angelo and Chen, C. L. Philip},
  title = {DCA-YOLO: A Dynamic Cross-Attention Network for Small Vehicle Detection in UAV Aerial Imagery},
  journal = {IEEE Transactions on Aerospace and Electronic Systems},
  year = {2026},
  volume = {62},
  pages = {15031--15047},
  doi = {10.1109/taes.2026.3715596}
}
Kostenlos starten →