Image surveillance systems are paid much attention by using embedded system gradually, but the most of they only focus on uncomplicated object tracking. If the system is not robust, it is easy to miss tracking after object interactions. This paper aims at this problem to propose the fast object tracking method based on texture and histogram techniques. An embedded surveillance system has been implemented on TITMS320DM6446 Davinci development kit. The inputs image sequence is from a CCD Camera, and the proposed algorithm is executed on the Davinci platform. The object tracking results are displayed on a 720x480 (NTSC D1) monitor with frame rate14 fps. From the experimental results, the developed embedded system can achieve better results in terms of functional practicality, tracking effect, efficiency, and computing power.