Research And Design Of Digital Video Surveillance System Based On DirectShow

Video surveillance is one of the main content in the field of Pattern recognition and computer vision, which has been widely applied in military, medicine and scientific research etc. The design of Moving target detection and tracking algorithm become a core in the video surveillance system. Therefore, the study of the key technology in video surveillance system and to improve the performance of the system are of great significance.Based on summarizing and analyzing the present technology, the thesis introduces the basis of image processing technology firstly, and then make further research and discussion on motion detection technology and object tracking.In image preprocessing, the thesis mainly introduces neighbour average filtering, median filtering, mathematical morphology, and so on. These technologies are used to remove the noise of the frame image and binary image.In motion detection, the thesis firstly introduces three algorithms that implement the motion detection, that is, frame difference, background difference, optical flow; and then analyze the principle of different kinds of algorithms, the advantages and disadvantages of different kinds of algorithms. On that basis, in order to improve efficiency of the system, the thesis proposes a hierarchical data structure based on regularly decomposed region quadtree. After that, the thesis uses horizontal projection and vertical projection to obtain the number of moving objects.In object tracking, the thesis mainly introduces four technologies which are, based on model tracking, based on region tracking, based on dynamic boundary tracking, based on feature tracking; and then mainly analyze the principle of the kalman algorithm. After that, in order to reduce search scope and improve efficiency, the thesis uses the kalman algorithm to predict the location for the moving object. For the multi-object, the thesis uses the moving object chain that contains the feature of the moving object. In the process of target matching, occlusion in image will influence the accuracy of the system, on this occasion, the thesis uses the predicted value to finish the target matching. After target matching, the thesis uses the observed value to update the parameters of the kalman filter and the object chain, then, the object tracking has finished.Finally, with the improved motion detection algorithm and tracking model based on Kalman filtering, the thesis uses MFC and DirectShow to implement the module of the system. Experiments prove that these methods are feasible and effective.