Object Tracking Based on Fragment Template and Multi-Feature Adaptive Fusion

Abstract

Object tracking under complex circumstances is a challenging task because of background interference, object deformation, obstacle occlusion, etc. Given such conditions, robustly detecting through single-feature representation are difficult tasks. For these problems, this paper presents object tracking based on a fragment and a multi-feature adaptive fusion. Through importing the concept of fragments, we distinguish the different types of occlusions, then adopt different the strategies of combining methods. Through importing the color, HOG and corner features, this paper also proposes a selfadaptive multi-feature fusion strategy based on their contributions. Experimental results show this algorithm can track moving objects robustly and accurately.