Abstract. In this paper, we propose the novel target tracking algorithm based on the visual saliency model. Dynamic object tracking problem of the scene is the current hot issues in the field of machine vision research. The main task of it from the dynamic target detection, recognition and tracking in image sequence, or even able to understand and describe the target object behavior that can be widely used in visual surveillance, human-computer interaction, robot soccer, military industry, etc. In view of the tracking object and the different application scenarios, multiple targets tracking problem has always been a difficulty in research of the target tracking. Tracking methods are divided into point tracking, nuclear track, contour tracking, and mixed scenarios. Our paper combines the concepts of the visual saliency to propose the enhance feature extraction methodologies that help to assist us get better tracking accuracy that is innovative.