Research Movement Of Human Visual Analysis

Human motion visual analysis technique belongs to computer vision [1] research field, the freeway, area residents, elevator, supermarket shopping malls and auto station, train station, airport and so on the occasions with safety requirements of high security intelligent visual surveillance system[2][3], can not only24hour surveillance target regional to promise security order problem, but also can make a large number of manpower and material resources to create and brought social undeniable economic benefits.This paper respectively experiments on a fixed background and no fixed background human motion video datas including cross jump, run and walk.After introducing difference detection method, background subtraction method, the optical flow detection method and Kim method,elaborating moving target detection algorithm principle, advantages and disadvantages and scope of application. For a fixed background data in the target foreground detection using background difference method, and without a fixed background in video in time difference method to detect.Having detected the foreground movement targets,two kinds of algorithm for motion feature extraction,one based on geometric moment and the other is thinning.Using the geometric moment feature extraction method is divided into contour-center method and the overall contour centroid method combined with regional moment looking for characteristic points method.In addition to the classic thinning algorithm based on index table,thinning algorithm is presented by using mathematical morphology convolution processing algorithm. In the end of explaining respective algorithm principle, their respective advantages and disadvantages, it carried out by frame difference with geometric feature extraction,background subtraction with geometric, feature extraction, regional feature extractionon the background difference, index-table thinning on background subtraction and the morphological thinning on background subtraction experiments.In the tracking section, introduced the Kaman filter and template matching tracking methods. After elaborating tracking algorithm theory,realization of the principle,core algorithm and the scope of application, combining the two into the experiments.All the experismental results in this paperare comes under VS2008development platform,using C,C++and VC++programming language, combined with OpenCV2.1source code library programming.