| Video surveillance system is developing toward the directions of audio/video digitization, system networking and case analysis intelligent up to today. In the course of the development of the video surveillance technology, the video surveillance system in the past twenty years generally experienced three generations-- the first generation is Analog video monitoring system, the second generation is Digital Video Surveillance System and the third generation is Distributed Video Surveillance System.The so-called intelligent video surveillance system is on the basis of digital, networked video surveillance ,but at the same time it is different from general video surveillance, it is a more high-end video surveillance applications. Intelligent Video Surveillance System is monitoring real-time detection of moving targets in the scene without any operators and then analysis, positioning, identification, tracking and behavior understanding the information. Strongly by the hand of computer data-processing functions, using image processing, pattern recognition and computer vision techniques, the intelligent video surveillance system can extract and analyze the video source key information, timely detect and handle the unusual cases under the monitoring scene, thus to be able to effectively assist the security personnel dealing with the crisis and maximize the reduction of false and omissions, so that it shows the surveillance system as the intelligent people. At our country to build a socialist harmonious society continuously push forward and make "building safe city, and promoting social harmony" as our society-wide goal, the intelligent video surveillance technology will be on a very broad prospects both in scientific research and engineering applications.This paper is written under such environment. After a detailed needs analysis, the goal of this paper is trying to design and achieve high scalability and application of intelligent video surveillance system from the application point of view, so the intelligent surveillance algorithm can be integrated into application-oriented systems in order to achieve a combination of theory and practice. Meanwhile, between algorithms and systems we can make the algorithm improvement and functions expansion separately without mutual influence. At the same time we can provide a complete development platform allowing the needs of future expansion and make it strongly reusable. Of course, evaluation of the merits or demerits of a system not only depends on the system architecture design, but also depends on the core algorithm design, the goal of this paper is to find a balance between algorithm complexity and system real-time. The design of algorithms request not only stronger robust and higher accuracy rate, but also lower algorithm complexity.Based on these two objectives, this paper firstly classified the functional requirements through the needs analysis, then made an overall structural design of the system, divided it into modules and get detailed analysis of each module to identify the implementation of key technologies of each module. Secondly, this paper made an in-depth research on these two key technologies of the moving target detection and tracking of the core algorithm, then put forward the methods of moving target detection and tracking which are used in this system with elaborate reasons and gave implementation flow. Lastly, this paper gave out other implementation methods of functional modules, and integrated the algorithms into the system to finally implemented the intelligent video surveillance system. The main content of the paper is as follows:1.Firstly, brief introduction of the course of development of video surveillance system, introduce the concept, research meaning and development at home and abroad of the intelligent video surveillance system, as well as the basic theory and technology which is needed by the system implementation.2.Discussing the key technologies of system application: DirectShow technology, MFC framework designing, video codec technology based on the FFmpeg, algorithm of moving targets detection and tracking. Designing the overall structure of intelligent video surveillance system which divides the system into four modules: video image capture module, video display and functions setting module, video recording module, intelligent video surveillance algorithm module.3.Detailed introduce and compare the current commonly used method of moving target detection. Because the mixed Gaussian background model can be updated online, is robust to the change scene, and algorithm complexity is not high, so the paper is focusing on the implementation process of mixed Gaussian background modeling used by the system, finish the code of the algorithm designing and implementing.4.Easy introduce the basic idea of various methods of tracking targets. For real-time consideration, the paper detailed describe the implementation process of mixed weighted tracking methods based on distance and color histogram which is used by the system, finish the code of the algorithm designing and implementing.5.Detailed describe the system's implementation process of each functional module: Implemented the video capture module based on DirectShow technology which can decode and capture the video files or real-time capture the image by video capture card. Implemented the restricted area alarm function based on the algorithm of moving target detection and tracking which can set the restricted area, when there is target moving into the area, system is alarming. Meanwhile finish the function of restricted area alarm log file recording, the system can record the log file when there is target moving into the area to facilitate searching the recorded video. Implementing the video recording module based on the FFmpeg which can code the video into H.264 and avi format. Implementing the main interface designing of intelligent video surveillance system based on VC6.0 development environment, the user can easily set each intelligent surveillance functions through the pop-up menu.6.Finally, the paper give the screenshots of system implementation, through the test the paper concluded: the system meets the design requirement, achieve the expected goal and operate stable, the algorithm accuracy and real-time is high.Intelligent Video Surveillance System covers a wide range of application, and the system's intelligent functions based on the algorithm of intelligent surveillance will be a lot. The system the paper introduced only implemented the basic functions of the intelligent surveillance, there are still a lot of work to finish in the future, summarized as follows:1.System is only for the application of intelligent surveillance, so there is no camera control part, system can be added the PTZ control module in the future for controlling the camera.2.System only implements one-channel video surveillance, it should be extended to four, eight or even more channels video surveillance in the future.3.At present the system is finished only in single computer, it doesn't add the function of multi-media information transmit in the network, the system should be expanded to adapt to the situation of network video surveillance in the future.4.System only implements the algorithm of moving target detection and tracking, and the restricted area alarming in various intelligent application, it can be integrated a lot of functions of intelligent surveillance in the future, for example: moving targets accelerate alarming, reverse alarming, people density estimating, target abnormal behavior analyzing, type of target identifying and so on.5.System should counterpoised the contradiction better between algorithm complexity and the real-time.As the future development trend of video surveillance system, implementing a intelligent video surveillance system of reliable and 24-hour real-time surveillance will have positive and far-reaching significance. Because different intelligent surveillance algorithm is suitable for different scenes, so the paper applied only to the surveillance scenes of single static camera and camera has a certain angle of depression and background is unchanged or changed slowly. The design of system can adapt to changes in algorithms as well as the expansion of functions, with in-depth study there must be algorithm which is suitable for different scenes in the future, then the scope of application will be more and more widely. |
PDF Full Text Request