|
|
|
| Adaptive Cross-Domain Fusion Method for Multimodal Heterogeneous Data in Battlefield Environments |
| TAO Qian1， PAN Jun2，3， LIU Hui4， YIN Qiang1， ZHOU Yu1， GU Jihong5 |
| 1. Chongqing Changan Wangjiang Industry Group Co.， Ltd.， Chongqing 401133， China； 2. Shanghai Electro-Mechanical Engineering Institute， Shanghai 201109， China； 3. Key Laboratory of Automatic Target Recognition （Shanghai ）， Shanghai 201109， China； 4. The First Military Representative Office of Army
Equipment Department in Chongqing， Chongqing 400042，China； 5. School of Integrated Circuits，
Nanjing University of Science and Technology， Nanjing 210094， Jiangsu， China |
|
|
|
|
Abstract To address the challenges of single-modality perception failure, high false alarm rates, and tracking discontinuity arising from the high maneuverability, severe interference, and frequent occlusion of low-slow-small unmanned aerial vehicles (UAVs) in complex battlefield environments, this paper proposes an adaptive cross-domain fusion method for multimodal heterogeneous data. The proposed method achieves spatiotemporal reference unification and cross-domain implicit feature alignment across heterogeneous data through an implicit feature alignment strategy leveraging three-dimensional occupancy grids and bird's eye view (BEV) representations. A Bayesian uncertainty-driven dynamic gating fusion mechanism enables adaptive anti-interference modality switching. Furthermore, by coupling radar micro-Doppler priors with a selective state space model, long-term trajectory closed-loop tracking under occlusion is realized. Experimental results demonstrate that the proposed method effectively overcomes the physical limitations of individual sensors, enhancing perceptual robustness and tracking continuity in highly adversarial, severely perturbed, and heavily occluded scenarios. This research offers significant theoretical insights and practical engineering value for advancing low-altitude defense equipment from multi-sensor hardware integration toward intelligent deep fusion, and from passive sensing to active collaboration, providing core algorithmic support for the development of all-weather, highly reliable counter-UAV operational systems.
|
|
Received: 26 March 2026
Published: 10 July 2026
|
|
|
|
|
|
|
|
|
2026, Vol. 9 
