Recognition And Tracking Of A Ground Moving Object Based On UAV

With the development of science and technology,the application of target recognition and tracking becomes widely used into various scenarios.It plays an important role in many fields,such as disaster relief,target-searching,video surveillance,etc.The traditional target recognition and tracking systems often use static video acquisition devices.They are composed of fixed installation of cameras.Thus,the scope of video collection is greatly limited,and it leads to the loss of the target easily in the process of target tracking tasks.In recent years,the rapid rise of the unmanned aerial vehicle industry has changed the whole information age.Nowadays,the civilian unmanned aerial vehicles have very good market prospect in many fields,such as aerial photography,geographical exploration,disaster rescue,circuit patrol,etc.As an indispensable part of the unmanned aerial vehicle,the camera device plays an irreplaceable role in the process of application.To overcome the flaw and insufficiency of the traditional target tracking systems,in this dissertation,a quadrotor was used as the platform for the target tracking system.We have studied three main parts of the related problems in depth: the ground moving target recognition and tracking algorithm,unmanned aerial vehicle localization method and target tracking strategy.Because the traditional target recognition and tracking algorithms are memoryless and easy to lose the target,we adopted a real-time online learning target tracking algorithm.This algorithm can timely track the target and update the feature of the target by means of the online learning method,so that it can adapt to the changes of the target status.In the part of UAV localization,a kind of monocular vision localization algorithm was improved to achieve accurate positioning of UAV.In order to make the target tracking system work stably for a long time,we presented a complete target tracking strategy.We divided the process of target tracking into three modes,including target-searching mode,target-tracking mode and target-lost mode.During each mode,a corresponding control strategy was designed to adjust the pose of the quatrotor to achieve real-time target tracking and get the target back when it was lost.The feasibility of the target tracking system was verified with a simulation platform.A series of experiments in the real environments were demonstrated to further verify the reliability of the target tracking algorithm and the completeness and real-time property of the target tracking strategy.