Multiple human tracking on image sequence under hierarchical attention control

The authors aim to implement flexible human visionlike processing on a computer by automatically switching the focus of attention to perform stable and efficient image sequence processing. To do so, they first introduce a hierarchical attention control model that divides high-order processing into five layers in this paper according to the object or role of each layer. By using this model, the authors implement attention control that uses not only simple image information, but also high-order information such as the various objects or relations between objects. Next, they take the problem of stably and efficiently tracking multiple people in a sequence of images to show the effectiveness of hierarchical attention control. In other words, by using this model, they determine the conditions of each person in an image based on the distance between the people or the way in which the people overlap and switch to suitable tracking processing or control the focus of attention. They use actual images to perform tracking experiments and verify the stability of tracking multiple people according to the proposed technique and the attention control processing contents of each layer during state transitions to show the effectiveness of the proposed technique.