Research on Background Learning Correlation Filtering Algorithm with Multi-Feature Fusion

Aiming at the problems of occlusion, drift, and background change in target tracking, a learning correlation filtering algorithm based on multi-feature fusion is proposed. In framework filtering, fusion, multi-template update, regularization are used to improve performance filter problem template contamination object occlusion. The fast directional gradient histogram (FHOG), color feature (CN), texture (ULBP) were extracted, channels connected series. Then depth features Conv4-4 Conv5-4 layers extracted through VGG-19 network, appearance model was constructed. To reduce sensitivity sudden background, constructed, alternate direction multiplier method (ADMM) speed up calculation filter. update stage, aiming pollution original caused by high-confidence strategy proposed fusing with highest confidence current frame, previous history frame. Finally, tested OTB50, OTB100, UAV123, TC128 experimental data sets, some classical latest algorithms. results show that tracking accuracy robustness improved.