Fig. 1. Global-aware object tracking framework

Fig. 2. Visualization of feature fusion

Fig. 3. Architecture of proposed global-aware network model

Fig. 4. Encoding-decoding module in self-attention mechanism

Fig. 5. Overall accuracy and success rate of different algorithms. (a) Precision; (b) success rate

Fig. 6. Precision of different attributes videos on PTB-TIR dataset. (a) Deformation; (b) occlusion; (c) scale variation; (d) background clutter; (e) low resolution; (f) fast motion; (g) motion blur; (h) out of view; (i) thermal crossover

Fig. 7. Success rates of different attributes videos on PTB-TIR dataset. (a) Deformation; (b) occlusion; (c) scale variation; (d) background clutter; (e) low resolution; (f) fast motion; (g) motion blur; (h) out of view; (i) thermal crossover

Fig. 8. Actual tracking results of six algorithms. (a) Airplane; (b) campus; (c) conversation; (d) road; (e) meetion; (f) crowd