Research On The Algorithm For Human Motion Detection And Tracking Based On Video Image

The detection and tracking of human motion target is key to the visual analysis ofhuman movement, which is also one of the important fields to computer vision research.It has wide application foreground and great economic value in many fields such as smartsecurity surveillance, advanced user interface, the details analysis of human movement,etc.Video Image sequences of moving Human can be divided into two types by in whichbackground is static or dynamic. In this paper we focus on the research of detection andtracking algorithms for moving human in the static background.In the Human moving target detection aspect，firstly, current theories of Humantarget detection are discussed, and analyzed their strengths and weakness. Then, theobjects detection methods based on Gaussian model are analyzed and discussed in detail.Taking their characteristics and applicable occasions into account, this paper usedbackground subtraction based on the adaptive Gauss background model to detect Humanmotion target. Because there are many noises in the video image, we must reduce themfirst and smooth image. The method of background subtraction is used to segment themoving target，and then remove shadow.In order to guarantee reliable motion detection,the background image is updated frequently.The small hole in the foreground can beeliminated by morphological filtering. The connected components labeling method isused to find the components which are connected in the binary image. Then the numbersand areas of the components can be obtained. If its area information and the ratio of theheight and the width can satisfy the characteristic of human, the component will beconsidered as a person. The experiments results show that the proposed method canobtain a more objects detection result.In the moving target tracking aspect, this paper summarizes Several kinds ofcommonly used methods，and analyzed the advantages and disadvantages of each method.Focus on the studies of tracking method based on Kalman filter and Camshift tracking method. The tracking method based on Kalman filter uses the motion information fortracking, but the Camshift tracking method uses the color information. On this basis thepaper presented an improved algorithm which combined Camshift algorithm with theKalman filter taking object moving and space color information into account.Successfully solving the interference problem from the color of Human moving objectwhich is similar to background and other objects or target is being covered. Theexperimental results show that the combination of the two algorithms not only reducesthe search time, but also improves the tracking accuracy and enhances the trackingperformance.