RGBT Object Tracking Algorithm Based on an Adaptive Feature Fusion Mechanism

To address the low accuracy and poor robustness of traditional RGBT object tracking algorithms, this paper proposes a new RGBT object tracking algorithm based on an adaptive feature fusion mechanism. Firstly, it enhances the feature extraction of each modality in complex scenes by introducing a spatial attention mechanism. Secondly, an inter-modal adaptive fusion mechanism is introduced to fully utilize the complementary information between the two modalities, enhancing the cross-modal fusion of RGB and infrared features. The algorithm is validated using authoritative object tracking datasets GTOT and RGBT234. The proposed method achieves highly competitive results and more comprehensive tracking performance, effectively solving the interference from similar objects and improving the tracker's robustness.