Self-Adaptive Visual Tracking Method for Illumination Varying Based on Multi-feature Fusion

This paper proposes a visual tracking method based on dynamic extracting multi-features to realize the robust and accurately visual tracking. First select the features that can compensate for each other to set up the feature mode using histogram. Second build the correlation function of local background illumination varying and dynamic amending features of target. Dynamic adjust feature set according to the variation of local background illumination using fisher criterion. Experiments and analysis done though the Particle-filter scheme show that this method can realize the robust visual tracking while environment of illumination varying and reduce the computational complexity.