Research On UAV Positioning And Positioning Based On Cooperative Target Under Monocular Vision

In recent years,unmanned aerial vehicles(UAVs)have been widely used in military and civilian fields,especially urban services,because of their compact structure,high flexibility and multi-dimensional information.Among these applications,the accurate acquisition of UAV's pose information is the premise of autonomous navigation,attitude control and other tasks.In this paper,a pre-calibrated camera with internal parameters is used as a visual sensor.Combined with digital image processing technology and camera pose estimation method,a set of solution for monocular vision UAV positioning and pose estimation based on cooperative target is designed.Addtionally,combined with uniform motion model,a motion estimation based on Extend Kalman filtering and visual pose estimation has been proposed.The main contents are as follows:Firstly,considering the distortion in real cameras,the second-order radial distortion model is studied.Using the calibrated camera distortion parameters to correct the extracted feature points.In this way,the non-linear camera pose estimation problem turned into a linear camera pose estimation problem.Experiments show that the accuracy of pose estimation is improved.Secondly,for the feature extraction of cooperative target,this paper designs a cooperative target with both color feature and point feature,and studies the algorithm of high efficiency image processing for detection of cooperative target,feature point extraction and matching.Compared with the high time-consuming problem of symmetric cooperative target matching problem,the cooperative target designed uses the geometric meaning of vector product and grayscale value can quickly get the pionts matched which significantly improve the efficiency of image processing.Besides,in order to get rid of the height limitation of visual estimation and improve the estimation accuracy by using multiple estimation center,this paper extended the basic cooperative targets to a distributed muti-cooperative target.Thirdly,in order to solve the pose estimation problem of UAV,the pose estimation algorithm based on weak perspective projection model is studied,and the pose estimation algorithm in perspective projection model is proposed.The orthogonal iterative algorithm is introduced to optimize the results.In addition,for the distributed cooperative object,a weighting method based on distance is proposed to improve the pose estimation accuracy.Finally,a motion estimation method combining the extended Kalman filter estimation and monocular vision pose estimation method,is proposed for the model-known motion or UAV motion smoothing.Using visual information as measurement equation,the pose prediction is realized by using EKF.Based on the uniform motion model,the simulation experiment is carried out for the points with noises.The result shows pose accuracy is superior to visual pose method used only under the same noise intensity.In this paper,the digital simulation experiments,the hardware experiments based on Vicon system and the experiments of four-rotor unmanned aerial vehicle platform have been conduted.The factors that influence the pose estimation are analyzed and the correctness of the algorithm is verified.Through the actual flight test of UAV,the correctness of the algorithm and the precision of pose estimation are verified,which lays a theoretical and technical foundation for further engineering application.