The Research And Application Of Human Body Detection And Tracking In Video

Since the beginning of the 21 century, the computer has almost related to all fields. At present, we human beings look forward to be able to use computers and related equipment to carry out a simulation of human visual. At the same time with the rapid development of multi-media technology in recent years, it is very convenient for us to capture video, which includes most motion information.Taking Video Image Sequence as the breakthrough point to explain and analyze the human action from different fields research directions such as Intelligent Monitor, Sports Video Analysis, Advanced Human-Computer Interaction, Virtual Reality, Image Coding Technology Based on Model and so on,which are the concern of Computer Vision.First, we describe the present methods of human body detection and tracking, and then we analyze and summary their strengths and weaknesses finally, Improved algorithm for human body detection and tracking is presented in this thesis. The three methods of human body detection (Background Subtraction, Frame Difference, Optical Flow) are illustrated in the Chapter 2. We have done a comparative analysis, their respective advantages and disadvantages are pointed out. We improve the combined Kim algorithm of Background subtraction and frame difference on the base of the above analysis. In the background modeling, we first adopt statistical average to obtain the background whether we have the previous background or not. Besides, this thesis takes account of the case that the gray-level of environment changes, and we adopt the Surendra method to update vision background. Then we add the results of the three frames difference Multiple to the difference between the current frame and the current background. Finally we could extract the better human body by Mathematical Morphology Methods. Meanwhile, we have done a vast amount of test work. The results indicate that the new algorithm has a good accuracy.In third and fourth chapter we focus on the description of the human tracking. Some representative tracking scheme, the essential technology, and the common prediction algorithm are summarized firstly.Especially,the Algorithm of Kalman filter which is adopted in our thesis is educed in detail.The basic flow of human body tracking based on the Algorithm of Kalman filter is discussed, and setting of its parameter is improved in order to improved the accuracy and reliability of tracking. Finally, we combine the centroid and area of human body and the Algorithm of Kalman filter and thus reduce matching time and finish the tracking. The simulation results show we could achieve the precision and robustness of tracking system, due to the previous better human detection.In the end, the conclusions of the thesis are reached, the existing deficiencies and the direction for further research are discussed.