A Multi-view Approach to Object Tracking in a Cluttered Scene Using Memory

In this paper, we propose a new multi-view approach to object tracking method that adapts itself to suddenly changing appearance. The proposed method is based on color-based particle filtering. A short-term memory and a global appearance memory are introduced to handle sudden appearance changes and occlusions of the object of interest in multi-camera environments. A new target model update method is implemented for multiple camera views. Our method is robust and versatile for a modest computational cost. Desirable tracking results are obtained.