Dual-branch Approach for Tracking UAVs with the Infrared and Inverted Infrared Image

The imaging for small unmanned aerial vehicles (UAVs) in infrared imaging systems is easily disturbed when they shuttle between different surrounding regions in a large background, resulting in target contrast inversion. Existing methods have poor tracking performance in this scenario. A dual-branch object tracking framework is proposed to tackle this problem. Firstly, the features of original and inverted infrared images are extracted by different branches. Then, a cross-correlation feature fusion module is proposed, which fuses the features of the two branches to enhance the features of the target area to obtain a more accurate bounding box regression. The proposed method achieves top performance compared with the existing state-of-the-art tracking algorithms, reaching the success rate of 78.39% and precision rate of 80.07% on the 2nd anti-UAV datasets, which are 5.84% and 4.52% higher than the SuperDiMP baseline.