Anti-occlusion Object Tracking Based on Multi-cue and Adaptive Particle Filter

摘要

Aimed at object tracking under occlusion, the traditional tracking algorithm could not work well especially when the object is similar to the background, an anti-occlusion object tracking algorithm based on multi-cue and adaptive particle filter is proposed. In the normal tracking mode, the method uses the color and texture features to present the object and multiplicative fusion strategy. When occlusion occurs, the dynamic modal is changed and the particles do only Brownian motion, weighted fusion, and then the particles are optimized by mean-shift algorithm to overcome the degeneracy problem and resume the tracking more quickly. Experimental results show that the proposed tracking algorithm is more robust and has good performance in complex scene.