Fragments-based object tracking using probabilistic graphical model

摘要

Occlusion is one of the most challenging problems in object tracking community. To deal with the occlusion problem, this paper presents a salient fragments based probabilistic graphical model (PGM) for visual tracking. In the proposed framework, salient points of visual object are firstly extracted, and then interest sub-regions or fragments of the object, derived from the salient points, can be obtained. Secondly, combining the feature information contained in each fragment and the spatial and temporal constraints between different fragments, the object is represented as a conditional random fields (CRF). Finally, based on the CRF model and Mean Shift tracking results of each fragment, a probabilistic inference scheme is adopted to estimate the object location. Comprehensive experiments on several testing videos show, compared with three well-known trackers, i.e. improved Mean Shift, Particle Filter and Fragments-based methods, the proposed method has a higher tracking accuracy and robustness, especially in occlusion condition.