Multi-Feature Fusion Target Re-Location Tracking Based on Correlation Filters

Target tracking has been a research hotspot in computer vision, and the correlation filtered target algorithm benefits of low computational complexity fast speed. Still, effect is not good when dealing with complicated circumstances. This paper proposes multi-feature fusion repositioning for problem complex environments. First, weighted presented. Since each feature different advantages environments, we combine HOG, CN, ULBP, image edge features use coefficient method to adaptively fuse component. Second, address occlusion problem, an judgment mechanism introduced, re-located by filtering. Third, scale pool established, filter trained classification search method. Finally, adaptive model update strategy proposed. We conduct comparison experiments current mainstream algorithms on publicly available datasets OTB-2015, VOT2018, UAV123, TColor-128, respectively, experimental results show that our proposed more robust scenarios.