A Learnable Spatial Mapping-Based Approach of Multimodal Feature Fusion Detection for Small UAV Targets in Complex Low-Altitude Urban Background
IEEE Sensors Journal · 2026-11 · 26 (21) · 33124-33143
DOI: 10.1109/jsen.2026.3735344 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Sensors Journal, 2026-11. Verfasst von Wen Jiang und 8 weiteren Autor:innen. Die Arbeit nennt 68 Literaturangaben. Gefördert von: Natural Science Foundation of China, (General Program), Beijing Natural Science Foundation.
Abstract (Originalsprache)
The current mainstream small uncrewed aerial vehicle (UAV) target detection methods still face several severe problems in complex low-altitude urban environments, including target occlusion, the similarity of thermal radiation characteristics relative to the surroundings, and insufficient significant features of small UAV targets, all of which result in a significant degradation of detection accuracy. To overcome these inherent technical limitations, this article proposes a novel learnable spatial mapping-based multimodal feature fusion detection (MFC-Det) method, which adaptively integrates complementary information from multiple heterogeneous sensing modalities. By performing precise learnable spatial alignment and joint feature optimization, MFC-Det effectively reinforces the critical target feature representation, achieving significant improvements in both detection accuracy and system robustness. The experimental results on our self-built dataset and the anti-UAV dataset show that the proposed method elevates the mean average precision (mAP) by 56.89% and significantly surpasses the existing leading approaches in detecting small UAV targets under complex low-altitude urban background conditions.
Förderung
- Natural Science Foundation of China
- (General Program)
- Beijing Natural Science Foundation
- Research and Development Program of Beijing Municipal Education Commission
- Youth Research Special Project of North China University of Technology
Zitieren
APA
Jiang, W., Zhang, K., Zhao, Q., Tian, Q., Wang, Y., Li, Y., Lin, Y., Bi, F., & Li, R. (2026). A Learnable Spatial Mapping-Based Approach of Multimodal Feature Fusion Detection for Small UAV Targets in Complex Low-Altitude Urban Background. IEEE Sensors Journal, 26(21), 33124-33143. https://doi.org/10.1109/jsen.2026.3735344
Vancouver
Jiang W, Zhang K, Zhao Q, Tian Q, Wang Y, Li Y, et al. A Learnable Spatial Mapping-Based Approach of Multimodal Feature Fusion Detection for Small UAV Targets in Complex Low-Altitude Urban Background. IEEE Sensors Journal. 2026;26(21):33124-33143. doi:10.1109/jsen.2026.3735344
BibTeX
@article{Jiang2026,
author = {Jiang, Wen and Zhang, Keyi and Zhao, Qiujie and Tian, Qing and Wang, Yanping and Li, Yang and Lin, Yun and Bi, Fukun and Li, Rui},
title = {A Learnable Spatial Mapping-Based Approach of Multimodal Feature Fusion Detection for Small UAV Targets in Complex Low-Altitude Urban Background},
journal = {IEEE Sensors Journal},
year = {2026},
volume = {26},
number = {21},
pages = {33124--33143},
doi = {10.1109/jsen.2026.3735344}
}
EUVIMED