Particle Filter Tracking Method Using Graphical Model
Mingjie Zhang, Baosheng Kang · 2016
To solve tracking problem when objects' part are temporally occluded in complicated scene, a novel particle filter tracking method using graphical model is proposed. The method applies particle filter to graphical models. When the color histogram of the tracked object is made, there is not one region of whole object but the multi-part region of it, thus it is divided into some parts, which means that the multi-part region is taken as a graphical model. In the process of object tracking by using particle filter method, messages about the state of the parts are sent to other parts in the graphical model. As the result, when the tracked object has occluded part, the proposed method can track stably and infer the state of the occluded part. Finally, experimental results verify the proposed method effectively.