Skip to Main Content
Occlusion is one of the challenging problems in object tracking, and plenty of tracking methods have been proposed to cope with this issue. Most of the methods deal with occlusion relying on observational or prior information of the tracked objects, such as appearance, shapes and motion. However, during occlusion especially serious and long-time occlusion, observations of object are hard to obtain, and prior knowledge, such as motion attributes, changes gradually over time. Therefore, modeling the object motion and then predicting the object's location until the object reappears, is likely to fail to serious and long-time occlusion. To cope with this problem, this paper proposes a novel method for object tracking with serious and long-time occlusion in image sequences based on occluder modeling. Occluder is modeled by detecting and evolving its rough partial contour represented by snake points, through minimizing the proposed energy function in which two novel terms are introduced: the push force and constraint force. Then, we search the tracked object around the neighborhood of the occluder contour until the object reappears. Experimental results demonstrate the effective performance of the proposed method on real sequences with total and long-time occlusions.