An improved multiple instance learning tracking algorithm based on occlusion detection

摘要

In this paper, we propose an algorithm to detect occlusion in visual tracking. Multiple instance learning is used to learn a discriminative appearance model for an object. When occlusion occurs, most tracking algorithms suffer from drifting problem for lack of effective occlusion handling. Therefore, in this paper, an occlusion detection module is added into the tracking framework, which is supported by a naive Bayes classifier. Moreover, the appearance model is updated according to the occlusion degree. Results of experiments conducted on several challenging video sequences demonstrate the superior performance of our proposed algorithm in terms of center location error and success rate of tracking.