Radar-guided Video Linkage Surveillance System And Algorithm Research

Video surveillance has been widely used in homeland security,coastal defense,civil and other fields.However,video surveillance system with camera as the core generally has problems,such as small monitoring scope,many blind areas and difficult target positioning and tracking,which obviously cannot meet the intelligent requirements of modern security monitoring.In order to realize the real-time positioning and tracking of video monitoring target,the thesis designed and realized the video linkage monitoring system,which was guided by radar with the ability of automatic target detection and intelligent positioning and tracking.A radar-guided video linkage surveillance model was established firstly in this thesis,and then a moving target location algorithm was proposed based on the model.According to the target information of radar real-time feedback,the corresponding camera azimuth and pitch angle of a moving target in the system linkage model were automatically calculated so that the target could be accurately locked,tracked and monitored in real-time.Aiming at the situation that multiple targets often appear simultaneously in a camera monitoring scene,an automatic multi-target selection algorithm was proposed which weighted fusion of the dispersion degree,approach speed and the distance between target and warning area,so that the target with the highest priority,namely the highest weight,could be monitored.In order to solve the problems of low target tracking accuracy,prone to deviation and loss in video monitoring,the thesis studied the principle,tracking process and characteristics of Mean Shift algorithm and Unscented Kalman Filter(UKF)algorithm,and proposed a target tracking algorithm combining UKF and Mean Shift.The algorithm took the position predicted by UKF algorithm as the initial position of the current frame search window,and then predicted and estimated the neighborhood of the initial position by Mean Shift algorithm.The experimental results showed that,the locating accuracy of the proposed target location algorithm for pedestrians and vehicles can achieve accurate target location.The proposed multi-target selection algorithm can effectively select the best monitoring target incomplex environment.The proposed target tracking algorithm has better automatic tracking effect than Mean Shift algorithm and UKF filtering algorithm,and has good accuracy,real-time performance and robustness.Finally,the thesis analyzed the function requirements of video monitoring system,and designed the architecture and monitoring process of radar-guided video linkage surveillance system with network video interface management technology,H.265 codec technology,streaming media transmission technology,etc.And it realized the function of data acquisition and storage,real-time preview,video playback,alarm,linkage monitoring interface and PTZ control interface,and verified the model and algorithms.This thesis,using characteristics that wide monitoring range and not affected by the conditions of optical of radar system,brokethrough the limitation of the camera optical visibility and visible region,which solved the problems of target positioning and tracking that results were not accurate and realized the all-weather,all-day,all-directional,long-distance video security monitoring.As a result of which,the linkage monitoring system can be widely applied to the places,such as border,oil field and coast,where have complex geographical position and a wide range of unattended area.And it can promote the security monitoring in the field of intelligent process,and laid a foundation for further application in the field of artificial intelligence technology in video monitoring.