Robust video object tracking algorithm based on multi-feature fusion
LI Yuan-zheng, Zhaoyang Lu, Jing Li · Journal of Xidian University · 2012
Object tracking using multiple features has poor performance under complex scenes and when occlusion occurs.Therefore,an algorithm for fusing multiple features adaptively in the particle filter tracking framework is proposed.The tracked object is represented by the fusion of all features under linear weighting,and a new method to estimate the fusion coefficient is also proposed according to the weight distribution of all particles as well as their spatial concentrations,thus improving the reliability of multi features fusion.Besides,a dynamic updating strategy is used to adjust the update speed of each feature template adaptively,thus alleviating the affection of object deformation.According to the confidence of each feature,an occlusion handling strategy is invoked to decrease the influence of partial occlusion.Analysis and experiment show that the proposed method is more robust under complex scenes,and is applicable in the presence of occlusions.