Research On Embedded Intelligent Video Surveillance System Based On Cortex-A9

People's big attention to the security?the disadva,ntages of high cost and low efficiency for traditional monitoring and the rapid development of network communication technology promote video surveillance towards intelligent and mobile direction.The difference between the intelligent video surveillance system and the traditional video surveillance system is:the intelligent video surveillance breaks the limitation of distance,and can realize the function of remote real-time monitoring;Combined with the image processing technology to realize the function of the moving object detection and alarm,target character recognition and tracking and so on.Starting from the development of video surveillance and the demand of intelligent surveillance system,this thesis focuses on the moving target detection algorithm emphatically and propose a algorithm that can detect more complete moving targets.The main contents of this thesis include:Firstly,the hardware and software platform is built.In terms of hardware,this thesis selects the iTOP-4412 development platform,it is equipped with Exynos4412 processor based on the Cortex-A9 architecture and the camera is ov5640 module.In terms of software,this thesis first ports Linux system to Arm development board and then completes the cross compiling and porting of OpenCV,at last,the camera driver is developed and transplanted according to the hardware information.Secondly,this thesis researches and realizes sub function modules that the video surveillance needed,including video capture,compression,transmission and display module.this thesis study the moving object detection algorithm and selects to combine the frame difference method with dynamic average background method on the basis of analyzing the advantages and disadvantages of them.The proposed algorithm solves the "holes" problem of frame difference method.For the selection of threshold that algorithm needed,this thesis adopt the adaptive threshold method,it can reduce the noise interference.Then,combined with the proposed algorithm,this thesis designs the overall implementation of system.The system includes two processes:real-time video surveillance and mobile detection alarm,and there are two threads in the process of real-time video surveillance.Two threads cooperate to complete the function of video compression and transmission.For the selection of threshold which is used to determine whether to alarm,this thesis through the experiment to determine the proportion of foreground pixels in two cases of the Invasion of characters and intrusion of smaller objects and sets it as 2%on the basis of analysis.With this threshold,when the moving target appears,the system can detects it accurately and triggers buzzer alarm,and at the same time,it can avoid the situation that the voice is identified as foreground target.At last,this thesis tests the whole system.The development board's Web site,the corresponding port number and the requested web page is input to the browser,then the monitor screen can be seen.When the moving targets appear,the buzzer is triggered immediately.