Figure 1.
Block diagram of long-term target tracking algorithm
Figure 2.
Template updater based on state judgment
Figure 3.
Confidence score chart
Figure 4.
SiamRPN tracking results
Figure 5.
Flow chart of target loss judgment mechanism
Figure 6.
Template-based re-detection algorithm
Figure 7.
Different ts corresponding success rates and accuracy
Figure 8.
Results of parameter optimization
Figure 9.
Diagram of LaSOT experimental results
TABLE I.
Experimental Results on VOT2018_LT
| Algorithm | F-value | Accuracy | Frame rate | Recall rate |
|---|---|---|---|---|
| SiamFC | 0.429 | 0.628 | 84 | 0.323 |
| MBMD | 0.613 | 0.636 | 4 | 0.576 |
| DASiamRPN_LT | 0.604 | 0.625 | 63 | 0.585 |
| SPLT | 0.614 | 0.629 | 26 | 0.602 |
| SiamRPN++ | 0.625 | 0.646 | 35 | 0.606 |
| Ours | 0.644 | 0.659 | 28 | 0.626 |
Figure 10.
Results of qualitative analysis of bird1 video sequence
TABLE II.
Introduction of 2 groups of video sequences
| Video Sources | Name | Number of frames | Type of challenge |
|---|---|---|---|
| VOT2018_LT | Yamaha | 3143 | Out of sight, occlusion, deformation |
| VOT2018_LT | bird1 | 2437 | Analogue interference, out of view, blocking |
Figure 11.
yamaha video sequence qualitative analysis results
Figure 12.
Ablation experiment results