A Hybrid Tracking Algorithm Based on Adaptive Fusion of Multiple Cues

摘要

A tracking algorithm based on adaptive fusion of multiple cues is proposed for moving target tracking under complex background. We compute the fusion weight adaptively according to the distinguish between target and background, which is measured by two-class variance ratio. Particle filer is used tr produce more samples at the case of multi-peaks and determines the final goal set position with the new particles. In addition, an adaptive updating mechanism is addressed to alleviate the mode drifts by the likelihood of the cues between two successive frames. Experimental results show the effectiveness of the proposed method.