Adaptive Confidence Map Fusion in Visual Object Tracking

Abstract

In this paper, we present a new multiple cues fusion algorithm in visual object tracking, which adaptive adjust the confidence scores based on center areas and surround areas defined on confidence maps. Confidence maps are created where each pixel indicates the probability of that pixel belonging to foreground object or scene background. Center areas and surround areas are used to calculate the confidence scores. The final confidence scores are created based on the calculation results and the old scores. Experiments show that the proposed algorithm has better results than traditional fusion algorithms.