Research And Design Of Intelligent Video Surveillance System Based On Moving Objects Detection Algorithm

With the development of network communications,computer vision,and microelectronics technologies,video surveillance systems are evolving toward intelligence,HD and networking.It has a very common application in various fields such as traffic,security,criminal investigation,and smart home.Video surveillance systems have become increasingly important in today's life.However,at present,most video surveillance systems still have some defects such as large size,wired connection,and monitoring by special person.The intelligence video surveillance system based on embedded platform not only can make up for these deficiencies,but also has attracted more and more attention of researchers because of its simple operation,high price/performance ratio,and strong computing power.Based on the above situations,an intelligent video surveillance system based on moving objects detection algorithm is researched and designed.The overall design framework of this system adopts the C/S architecture,in which the server is based on the embedded development board X210 and its peripherals(mainly including USB camera and wireless network card).It uses Linux as the operating system,and its core processor is Samsung's S5PV210.In order to facilitate the development of graphical interface and the realization of the corresponding function ported Qt library on the server.A Windows 7 system PC is achieved as client.The design goal of the entire system: the server captures video data through a USB camera and displays on the local LCD display in real time.At the same time,using multi-thread technology,the video data is compressed in the H.264 standard and transmitted to the client through the network in the form of data streams with RTP/RTCP network protocol.The client receives and decodes the data streams,and displays on PC.At the same time,the improved algorithm proposed in this thesis is used to detect the moving objects of the video.Finally the more completed moving objects are detected on the client.In terms of the improved algorithm used in the systems,after studying the existing algorithms used for objects detection,the Vibe is selected for excellent features such as small calculation amount,small occupied space,high processing speed,and high detection accuracy to further optimization.There are three shortcomings of Vibe: eliminating ghost slowly,generation holes and small objects easily and sensitivity to shadowed areas.The algorithm is improved in these three aspects.Firstly,the reasons why the Vibe can detect ghosts are analyzed.For the feature of the ghosts that different from the moving objects is the positions of the continuous frame image not changed.Therefore,the first successive few frames are differentiated,and the ghosts can be positioned quickly,thereby elimination of ghosts also quickly.Then,referring to the idea of impfill algorithm,using the combination of set operations and floodfill function,the methods of processing the small objects are added simultaneously with the processing of the holes,and the accuracy of the detection results is improved.Finally,the Seed Fill algorithm is improved by analyzing the pixel connectivity features of the shadow regions to find the connected regions,which applies to the detect shadow regions.The robustness of the improved algorithm in detecting moving objects is improved.Using several sets of image sequences(video)in the ChangeDetection data set as input data for experimental comparative analysis,the results show that the detection results of the improved algorithm are more accurate.In addition,the improved algorithm of this thesis performs better than the traditional Vibe algorithm in comparison with data such as accuracy rate and recall rate.Finally,the entire video surveillance system is implemented.The codes of each module are transplanted in a built platform and environment to achieve the desired goal: video capture,video display,data compression and encoding,data transmission,etc.on the server;and video reception,video display and detection moving objects on the client.The test of the constructed system has basically reached the expected design goal.Through the cooperation of various functional modules,video data is captured and displayed.Moreover,the moving objects in the video can be detected more accurately.The research and design of the intelligent video surveillance system based on moving objects detection is impleted.