Research On Target Recognition And Tracking Methods Of Armored Vehicles

Image identification and tracking technology has long been the hotspots in the domestic and international research field.It has been widely applied in Aerospace,security surveillance,military and other fields.The algorithm design of target detection and tracking directly affects the accuracy and stability of tracking effect.Aimed at armored vehicle targets in video sequences,this thesis focuses on the research in the recognition and tracking algorithm.Firstly,the features of the AdaBoost algorithm,weak classifier and strong classifier are summarized.After studying the AdaBoost algorithm in depth,the relevant positive and negative sample database for the armored vehicle target is established,and the strong classifier for the target recognition through the learning training method is obtained,and these can effectively help identifying the target in the static background.In this thesis,several classic target detection algorithms are analyzed,and the advantages and disadvantages of the algorithms are analyzed by experiments,and the existing algorithms are modified and adjusted for the experiments.A method of combining Gaussian background modeling with differential method is proposed,which combines the advantages of the two methods and can detect the moving target rapidly and adaptively.In this thesis the typical tracking algorithms are analyzed and then the research focuses on algorithms based on Meanshift and Camshift about moving target tracking.A series of improvements are conducted on the shortcomings of the algorithm of the Camshift which use the result of the moving target detection as the tracking motion region and this can track targets automatically.Meanwhile,the Camshift algorithm search window is adjusted and the dynamic stability search window is used to solve the problem of the persistent stability in the target tracking.The Kalman filter algorithm is introduced for the problem case of occlusion,color interference and other moving objects,combined with the results of strong classifier recognition can recognize and track the armored vehicles more effectively.Finally the future work and the prospect of the research field of identification and tracking algorithms are discussed,the problem about tracking in complex variable environmental and multi-target identification should be further studied in the future.