The Algorithm Of An Improved Moving Target Detection And Its Application In Video Surveillance System

The video surveillance system is a combination of computer vision, imageprocessing, pattern recognition, artificial intelligence, and digital signal processingand other disciplines. Video information provided by the traditional video surveillancesystems are basically without any analysis, and during the monitoring processes, itneed monitoring personnel make some operations. For example, when an exception ordanger occurs in the scene, it need human subjective analysis and judgment thesemonitoring information, so, this monitoring system wastes a lot of manpower. Withthe development of science technology and image processing technology, videosurveillance systems have been expanded, therefore, the traditional video surveillancesystems which rely solely on manpower cannot meet the urgent needs of people andlife, video surveillance systems with automatic or semi-automatic monitoringfunctions become the future development trends of the video surveillance system. Itsmain research contents include extraction, description and tracking of video movingobject and target identification and analysis, and it has been widely used in publicsafety monitoring, traffic violations, traffic monitoring, hospital care and other fields,and it has far-reaching theoretical and application prospects.Moving target detection technology is one of the most basic skills in the study ofvideo surveillance systems, and its test results directly related to the subsequentpositioning of the moving target identify and track. The domestic and foreign researchinstitutions and corporate researchers have invested a lot of manpower, material andfinancial resources to study it. The current typical moving target detection algorithmcontains: Optical flow, background subtraction method and intra-frame subtraction,and they each have advantages and disadvantages. Based on camera fixed, detectingtarget movement as the research scene, this paper studies the main problem of movingtarget detection algorithm, and improve the existing algorithm, using the USB cameraas the system of video acquisition unit, visual c++and OpenCV6.0as thedevelopment platform program, design and realization a video monitoring system with automatic alarm function.The main researches in this paper as follows: Firstly, introduce this researchbackground and significance of the topic, brief introduction of the course ofdevelopment of the video surveillance system, as well as development at home andabroad. And then details introduce the image processing techniques: RGB, HSV andYUV color model, and image pre-processing technology, as well as the currentlypopular image compression algorithm. Then through the research on the needsanalysis of video surveillance system, classified the functional requirements, andmade an overall structural design of the system, divided it into some modules and getdetailed analysis of each module to identify the implementation of key technologies ofeach module. After that, detailed introduction and compare the current commonlyused method of moving target detection: Optical flow, background subtraction method,intra-frame subtraction, and focus on frame difference method and backgroundsubtraction method. Based on the existing algorithms, by combining the adaptabilityof the frame difference method and the accurate of background difference method, amoving target detection algorithm based on background subtraction method combinedand mixed-frame difference frame difference has been proposed, to improve the"empty" and "ghosting" phenomenon and achieve the rapid detection of Movingtargets. Finally, the paper gives the screenshots of the system implementation, throughthe test the paper concluded that the system meets the design requirement, andachieves the expected goal.However, there still needs further work of this paper, for example, it is mainlyfocus on researching the indoor situation of the camera head fixed and the backgroundenvironment still, so the system applies limited occasions, further research will beadded the control of the camera, or extended to multi-channel video surveillance.