On-line Discriminative Feature Selection in Particle Filter Tracking

This paper presents a particle filter for object tracking using the combination of shape and texture features. Local descriptors contribute to estimation by filtering out some irrelevant observations, making it more reliable. We introduces an online feature adaptation mechanism that enables to automatically select the best set of features in presence of time varying and complex background, occlusions, etc. Experimental results on real-would videos demonstrate the effectiveness of the proposed algorithm.