Fig. 1. Tracking images in different states and corresponding foreground and background color probability maps. (a) Original image; (b) foreground color probability map; (c) background color probability map

Fig. 2. Logarithmic loss function graph

Fig. 3. Partial tracking framework

Fig. 4. Target and response result graph. (a) Normal tracking; (b) response map under normal tracking; (c) background clutter; (d) response map under background clutter

Fig. 5. Schematic diagram of HCAF algorithm framework

Fig. 6. Precision and success rates of occlusion attributes on OTB100 dataset

Fig. 7. Precision and success rates of background clutter attributes on OTB100 dataset

Fig. 8. Precision and success rates on OTB100 dataset

Fig. 9. Precision and success rates on LaSOT dataset

Fig. 10. Tracking results of 10 tracking algorithms in partial sequences. (a) Basketball; (b) Human3; (c) Jogging-1; (d) Soccer