Number of approaches has been proposed for object tracking in last few years towards increasing performance of computer vision systems. In this paper we propose a tracker which is based on structural information captured in pixels, thus we facilitate a tracker which distinguishes between the target and the background easily. This target-background posterior estimate is further used to track the object. It is used for effective tracking of objects in generic computer vision system.
It works in heavy occlusion and illumination conditions also. The experimental result shows the effectiveness and success of the tracker during heavy occlusion also it recovers the image sequences from drifts. The approach includes using k-means for creating clusters in order to form the model based on the pixels. The results are encouraging. Furthermore, the proposed algorithm segments the target and background during tracking the objects.