Exploring fusion strategies for accurate RGBT visual object tracking

We address the problem of multi-modal object tracking in video and explore various options available for fusing complementary information conveyed by visible (RGB) thermal infrared (TIR) modalities, including pixel-level, feature-level decision-level fusion. Specifically, contrast to existing approaches, we propose develop paradigm combining image fusion at pixel level. At feature level, two different kinds strategies are investigated completeness, i.e., attention-based online strategy offline-trained block. decision a novel is put forward, inspired success simple averaging configuration which has shown so much promise. The effectiveness proposed owes number innovative contributions, dynamic weighting RGB TIR contributions linear template update operation. A variant method produced winning tracker Visual Object Tracking Challenge 2020 (VOT-RGBT2020). comprehensive comparison with highlights advantages multimodal score Extensive experimental results on five challenging datasets, GTOT, VOT-RGBT2019, RGBT234, LasHeR VOT-RGBT2020, demonstrate robustness method, compared state-of-the-art approaches. Code https://github.com/Zhangyong-Tang/DFAT.