Research On Moving Object Detection And Clarity Method In Video Surveillance

With the increase of vehicles and traffic intensity,the traffic jam has become a serious problem that challenges the development of the cities.Intelligent transportation system,which shows great effectiveness in solving the traffic jam and improving traffic efficiency,has got more and more attention.In such system,the technique of video image detection and Clarity Processing in traffic monitoring plays a key role.In this paper,we give some insights into the research on the traffic intersection surveillance technique based on vision,which includes detection,extraction and Clarity Processing under complicated scene.The main work of the paper is as follows:(1) Analyze and compare the advantages and disadvantages of the current classic algorithms of the moving object detection,and on this basis a modified moving object detection algorithm of optical flow based on temporal difference is proposed. According to the characteristics that the Optical Flow is sensitive to velocity,the Lucas-kanade algorithm is combined with the Gaussian Pyramid Layered algorithm,so it has good stability and noise immunity.This method can effectively extract the moving object from the urban transportation videos.(2) In the face of the clear processing about the moving object,the reason why the video image in the traffic control system is not clear is analyzed.On this basis,the constrained least square method is applied to remove the motion blur,and the Kalman filter is applied in the clarity processing to eliminate the noise blur.Aiming at the complex motion scenes in the intelligent transportation,an algorithm of video image clear processing based on the moving object detection is proposed to use the moving object obtained from optical flow based on temporal difference detection to guide the clear processing of video image,that is,only the interested regions are processed to reduce the computation and improve the real-time performance.(3) On the basis of the above work,the algorithm in the paper is implemented with MATLAB and OpenCV,and a prototype system about Clarity processing of video image in traffic control is designed and developed.The several groups of traffic video files are collected to do the experiment which shows that the algorithm in this paper can process the moving object in traffic scenes accurately and efficiently with certain of robustness.