A real-time tracking method based on SURF

Two important problems in real-time tracking are: 1) how to discriminate an object from clutter environments, and 2) how to meet the real-time requirement in practical applications. In real-time tracking application scenario, neither priori information about targets nor background model is known, which makes many traditional methods fail. In this paper, we propose an object detection and tracking method based on Speeded-Up Robust Features (SURF). A region of interests is set up to reduce computation burden and an adaptive reference library is built and updated by reusing the extracted feature points and past object location. The advantages of this method lies in its robustness while its calculation is light. Our experiments show that our method is robust under camera wobble, background clutter and illumination changes. It can reach real-time processing in various occasions.