Autonomous Tracking And Landing Method Of UAV With Monocular Vision For Ground Mobile Robot

Computer vision is an important auxiliary technology in the unmanned aerial vehicle(UAV)applications.It has been widely used in UAV applications such as military detection,police patrol,aerial photography and electric power inspection.The technology has the advantages of large amount of information,high independence and good real-time.It can help UAV get rich external information and help UAV to run more efficiently and more accurately.Therefore,the research of UAV system using visual technology assisted navigation is of great significance.When UAV performs the task,it can be divided into two sub tasks:independent flight with reference task path and autonomous landing after task completion.In two sub tasks,the introduction of visual auxiliary technology greatly improves the flight performance of UAV.In this paper,the autonomous tracking and landing of UAV on monocular vision for ground mobile robots was studied from the following aspects:Firstly,the autonomous tracking and landing mission system of UAV equipped with monocular camera for ground mobile robot was built.The DJI M100 UAV was selected as the flight platform,with sensors such as GPS and Guidance.The Nomachine software was used to control the UAV in connection with the data link between the on-board computer and the ground station computer.Using the tensor calibration method to calibrate the single camera on UAV.obtain the internal reference of the single camera and correct the image distortion of the single camera,and help the subsequent image processing algorithm to get more accurate calculation results.Secondly,the physical modeling of UAV was carried out,and the control principle of various flight states of UAV was analyzed.The conversion between common UAV coordinate systems was deduced to solve the problem of GPS signal conversion in UAV navigation.Based on the DJI Onboard SDK software development platform of DJI M100 and fusion of GPS and Guidance sensor data,the autonomous flight algorithm was written under Linux system,and the experiment was designed to test and verify the autonomous flight algorithm of the system.Thirdly,the control method of the UAV vision assisted autonomous landing and the image recognition algorithm on the mobile robot were studied,and the autonomous tracking of the UAV to the ground mobile robot was realized.In the autonomous landing of the UAV on the ground mobile platform,the two-dimensional code identification of the combined scale and the motion estimator of the particle filter were designed to solve the problem of the lost target image in the camera field of vision.Finally,the feasibility of the algorithm was verified by experiments.When autonomous control was carried out by UAV,visual identification of two-dimensional code and visual feedback in horizontal direction are adopted to realize the tracking of ground mobile robot.On this basis,keeping the tracking effect in the horizontal direction,adding the vertical direction control of the UAV,gradually reducing the altitude of the flight,and realizing the autonomous landing function of the UAV on the ground mobile robot platform.Through experiments,the UAV could achieve high accuracy of fixed-point hovering,track the ground mobile robot stably and achieve accurate landing.