Research On New Tracking Algorithm Of Mean Shift For Intelligence Surveillance

In recent years, with the development of computer skills and computer vision, intelligent video surveillance system has become a new application direction and a cutting-edge issue of concern. Moving object tracking based on image sequences is one of important research topics in the domain of computer vision and digital image processing. In this thesis, with intelligent video surveillance system as the main research background, under the Mean Shift algorithm framework, problems of feature selection and the change of target appearance were researched, and with the above as base, an automatic and relatively complete moving object tracking system was constructed. The main contents of this thesis as follows:(1) In this thesis, the target tracking problems which based on the original Mean Shift algorithms are studied firstly, and the experimental results verify that the original Mean Shift algorithm could make the target tracking failure when the color of background and object is similar. On this basis, an improved tracking algorithm is proposed(MSL). That is, apart from the conventional color histogram features, the texture features of the object are also extracted by using the local binary pattern (LBP) technique to represent the object. The major uniform LBP patterns are exploited to form a mask for joint color-texture feature selection. Compared with the traditional color histogram based algorithms that use the whole target region for tracking, the proposed algorithm extracts effectively the edge and corner features in the target region, which characterize better and represent more robustly the target. The experimental results validate that the proposed method improves greatly the tracking accuracy and efficiency with fewer mean shift iterations than standard mean shift tracking. It can robustly track the target under complex scenes, such as similar target and background appearance, on which the traditional color based schemes may fail to track.(2) In the view of problems of conventional target tracking algorithms, a novel template image updating strategy based on information of gray histogram is presented. In addition, a track prediction method based on kalman filter is proposed to track the target reliably when the target is occluded badly. In the real target systems, the target move approximately at uniform linear motion in local period, for which a simple and efficient track prediction method is proposed. The experimental results show the proposed method has the ability of predicting the track of target rightly when the target is occluded badly. In addition, the proposed template image updating strategy promotes the reliability of system.(3) Based on the object tracking algorithm, using VC++development platform and OpenCV, a simple moving object tracking system based on video was designed, and through this system verified the feasibility and practicability of the algorithm.In this thesis, Basic Mean Shift tracking algorithm has been researched and improved from feature selection and template image updating.The changing algorithm can track object well even when the object’s color is similar to the background or the appearance of the target changes. At last, a practical automatic moving object tracking system was constructed.