Intelligent Monitoring System Based On Video Analysis And Realization

As there are some disadvantage on the traditional video surveillance system, such as single function, large storage space occupation, running low efficiency. To meet the market requirements, enhance product competitiveness, it's necessary to improve the design of original video surveillance system, to achieve intelligent video surveillance by the powerful computer processing ability.This subject locates on the workflow of intelligent video surveillance system processes, study the way of video data importing, video data analysis, incident response, etc.For the video data importing, the important point is study the video data acquisition methods of the WDM device and IP camera. To WDM device, the video data can be imported by DirectShow for saved as a file or introducing the next step of the analysis; To IP cameras which support MJPEG format, it's necessary to interact with the camera by HTTP protocol, for the response in MJPEG stream, need to split the stream data to single frame in JPEG format by delimiter. A unified interface and plug-in framework was used in code implementation for the expansibility of the program. For the analysis of video data, the main research scopes are implementation and suited scenario of the Frame Difference method, Background Difference method and other methods. The two frame difference method and the three frame difference method were compared, the conclusion to a three frame difference solves the ghosting problem. Background subtraction to achieve given the background during the movement of objects the actual verification, achieved good results, The main idea for obtain the background image is get it from background training set as Gaussian model. To facilitate monitoring, the program improved the method of moving object mark, achieved good visual effects, and the program supports the statistics the number of moving objects and sports aspect ratio the portion of the statistics. Image filters in AForge framework was used to complete the complex image processing. For the incident response processing part, the major content of study is the approach of save video to storage and implementation of SMS notification, and provided some idea for other incident response actor. Incident response module is also following a plug-in framework, to facilitate future expansion of the program features.