Real-Time Compressive Tracking with a Particle Filter Framework

Recently a real-time compressive tracking was proposed and achieved relative good results in terms of efficiency, accuracy and robustness. It belongs to the "tracking by detection" method. Slight inaccuracies in the tracker can lead to incorrectly labeled training examples in these algorithms, which degrade the classifier and usually cause drift. In this paper, we incorporate the motion model into the traditional compressive tracking where we utilize the particle filter. Therefore, our algorithm can handle drifting problem to some extent. Meanwhile, in order to improve the discriminative power of the classifier to relieve drifting problem radically, a modified naive Bayes classifier is proposed. The proposed algorithm performs favorably against state-of-the-art algorithms on some challenging video sequences.