Research On Image Tracking Of Ground Moving Object

Research on image tracking of ground moving target refers to the technology of detecting,identifying,locating and tracking the target from video sequences,and has a wide application in monitoring and unmanned aerial vehicle tracking.The technology can detect the azimuth,speed and other information of the ground moving target in real time,and also can carry on the behavior analysis,the monitoring and more advanced tasks.However,in some cases,due to the background movement caused by the movement of the camera makes the traditional method is difficult to implement,which brought great challenges to the late target tracking.The basic knowledge of image processing,target detection and tracking are studied systematically.The background compensation method of gray scale projection is studied in detail.Based on the analysis of various algorithms,the basic idea of improvement is:(1)The least squares method is applied to the gray scale projection algorithm,and the 2-point linear approximation prediction mechanism is applied to the next frame,and then the background displacement to be predicted,thereby improving operational efficiency.In the aspect of the detection and prediction of background displacement,due to the unimodality of the correlation function,the idea of folding the search interval is introduced,which greatly improves the efficiency of the algorithm;(2)A five-frame difference method and background subtraction combined with a canny operator are proposed for the problems such as the unclear contours,the similarity between the target gray scale and the background gray scale,and the difficulty of controlling the motion,the method can get the precise area of the moving target,so that the target area contour is more clear,increase the accuracy of detection;(3)Aiming at the shortcomings in the traditional target tracking algorithm,such as the Mean-shift algorithm can not track the target of occlusion and size change.The Camshift algorithm can not resist the interference of similar color objects,The particle filter algorithm is high complexity,time-consuming and has particle degradation.problem.Therefore,this paper proposes a combination of the meanshift algorithm and the particle filter algorithm to converge the particles using the clustering of the meanshift algorithm so that each particle converges toward the vicinity of the moving target,which greatly reduces the number of particles needed to desctiibe the target state information number,shorten the time consumption,to achieve realtime accurate tracking purposes.