Efficient Visual Object Tracking with Online Nearest Neighbor Classifier

A tracking-by-detection framework is proposed that combines nearest-neighbor classification of bags of features, efficient subwindow search, and a novel feature selection and pruning method to achieve stability and plasticity in tracking targets of changing appearance. Experiments show that near-frame-rate performance is achieved (sans feature detection), and that the state of the art is improved in terms of handling occlusions, clutter, changes of scale, and of appearance. A theoretical analysis shows why nearest neighbor works better than more sophisticated classifiers in the context of tracking.