The Research Of Tracking Motion Human Based On Meanshift Algorithm

Human tracking is one of the most important topics on visual analysis of human motion. Since its widespread application prospects and potential economic value in intelligent monitoring area, it has inspired the interest of many scientific researchers. Tracking motion human has become one of the hot points of research.This paper at first summarizes the background and current situation of visual analysis of human motion research. And then it describes the practical significance of application in intelligent monitoring area. Moreover, the advantages and disadvantages of a variety of human tracking methods are analyzed. And the difficult issues on human tracking are also discussed.Meanshift tracking algorithm and other algorithms which are improved on robustness are introduced as following. On the basis of above, a new improved method is put forward to solve the problem of lacking discrimination effect and stability in Emilio Maggio and Andrea Cavallaro's regional blocks method. It combined with human color distribution expects to raise the robustness of human tracking under complex monitoring environment. On one hand, it reduces processing time by cutting down the number of human regional blocks, yet not losing related spatial information. On the other hand, adding weighted coefficients for each block improves the discrimination effect. In addition, the problem of tracking human under completely occlusion in monitoring is resolved. The improved tracking algorithm combined with kalman filter can obtain better tracking results under linear motion occlusion circumstances. But it is unsuitable for non-linear motion occlusion situations. Making use of the advantage of integral histogram method can complete exhaustive search fast. So tracking algorithm combined with integral histogram can get relatively better tracking results under both linear and non-linear motion occlusion situations, increasing the robustness of tracking human.