Design And Implementation Of Human Target Detection And Tracking System Based On The Real-time Surveillance Video

With the rapid development of social economy and urbanization, as well as theimprovement of the image acquisition of computer technology, video surveillance hasgot rapid developed and been widely used in various fields, which plays a vital role infighting terrorism and building a wisdom safe city and also become a criticalcomponent of the social security system. However, the presence of some restrictionfactors in traditional video surveillance has brought some problems, such as lowinefficiency, the waste of human and financial resources. Thus, the intellectuality ofvideo surveillance has become an inevitable development trend.Aiming at two key components of the intelligent monitoring, which are movingtarget detection and moving target tracking, a human target detection and trackingsystem based on real-time surveillance video is designed and implemented fromhardware and software in this dissertation, through analyzing existing imagepre-processing methods and the human target detection and tracking algorithms. Forthe hardware part, a CCD camera combined with a digital image acquisition card withPCI interface and a PC is utilized to capture, input, store, process and display thedigital video images. For the software part, the real-time color video image sequencescaptured by CCD camera is pretreated by graying, smoothing and binarization usingVisual C++6.0on Microsoft Windows XP32-bit operating system platform, wherethe algorithm is achieved with C/C++programming language. The foreground objectis obtained using background subtraction method based on Gaussian mixturebackground modeling. Besides, removing disturbed area and exacting completeforeground target area is necessary, which is achieved by employing morphologicalprocessing and connected area analysis method. Then, the recognition of humantargets is realized combined with the shape feature of the exacted target area. Theapproach of Kalman filtering and minimum enclosing rectangle of the target isutilized to predict the position of human target in next frame, thus to achieve the goalof human target tracking.