A Novel object Tracking Algorithm Based on mean shift algorithm and SURF

Abstract

Mean shift algorithm (MSA) is a powerful object tracking technique due to its simplicity and robustness. However, it causes easily the inaccurate tracking when the scene is complex, for example, the object is seriously occluded, the color of the object is similar to that of the background, the background is dynamic, and the camera shakes or moves. Aiming at the above problems, a novel object tracking algorithm based on MSA and SURF (Speeded Up Robust Features) is proposed. Firstly, the object area in the current frame is determined by MSA, and secondly, SURF algorithm is used to match the feature points in the object area of initial frame with those in the object area of current frame, and finally, the coordinates of the center point in the object area of current frame are adjusted according to the matching results. The experimental results on three videos in complex scenes show that the proposed algorithm can track the object more accurately than MSA, and realize the real-time and accurate tracking of video objects in complex scenes.