A Robust Tracking Algorithm Based on Feature Fusion and Occlusion Judgment

Abstract

How to achieve a robust performance remains an intractable problem in the various object tracking algorithms due to some unfavorable factors, e.g. occlusions, appearance change, etc. In this paper, a robust object tracking approach is proposed based on feature fusion and occlusion detection. Under the relevant filtering model, two complementary features, HOG and color name features, are fused via a weighting strategy. Moreover, an occlusion detection method is presented according to the response function of the fused features. Experimental results on several challenging sequences demonstrate the effectiveness and feasibility of the proposed method.