Research On Target Tracking And Localization Algorithm Of UAV Based On Computer Vision

With the rapid development of computer technology and sensor technology since1980 s,The research of unmanned aerial vehicle(UAV)has also made a lot of achievements in many countries' research institutions.UAV equipped with visual sensing equipment can not only achieve the function of aerial photography,but also can achieve the detection and tracking of specific targets.Unmanned aerial vehicle(UAV)for moving target tracking has important theory significance and broad application prospects.Understanding the control principle of UAV and the processing of visual information is the basis for the detection and tracking of specific targets.How to deal with the visual information to achieve the robustness of tracking is the focus of current research.In this paper,quadrotor UAV aircraft as a platform for the target will be lost in traditional computer vision tracking,and when tracking the environment background color and tracking similar color objects tracking failure when easy problems are analyzed.Firstly,the control principle of the quadrotor unmanned aerial vehicle(UAV)platform is established,and the coordinate transformation matrix of the four rotor aircraft is analyzed.Secondly,the visual images acquired by the quadrotor UAV equipped with vision sensing devices are susceptible to the changes of illumination and noise,so the basic image processing methods are analyzed.On this basis,the target detection method of frame difference,optical flow and background difference are analyzed and tested.According to the above detection method using conditions and disadvantages,this paper proposes a visual information detection method combined with Opencv.We collect the RGB image after color space conversion into a HSV image which is little affected by illumination change.The image noise removal and morphological filter,this method can effectively detect moving targets outline.Then the basic principles of Meanshfit,Camshfit and Kalman tracking methods are analyzed.Camshfit has a great influence on the illumination during the tracking process and kalman filter is difficult to be applied to nonlinear systems.Using opencv of image denoising for target tracking in the video images.Thendetecting and finding the largest contour of targets in images and throughing the IMM kalman filter to modify the target location.Calculating the quadrotor aircraft flying toward to achieve quadrotor aircraft in the area of video moving target tracking.Finally,the tracking system platform is built by using AR.drone four rotor and Opencv.With the help of Visual Studio 2013 software,the detection and tracking of moving targets are carried out.The experiments show that the proposed method can effectively realize the accuracy,stability and rapidity of moving target tracking.