Position And Attitude Measurement Based On Monocular Vision

The position and attitude measurement of unmanned aerial vehicle(UAV)is an important problem in modern UAV control,navigation,tracking and many other fields.The position and attitude information of UAV is an important parameter to reflect its motion state.It is of great value to obtain these parameters for UAV flight control.Computer vision measurement technology,which has the advantages of non-contact measurement,signal acquisition,data processing and high automation,has been applied to UAV pose and attitude measurement.The main task of this dissertation is to use monocular vision imaging technique,CamShift encoding algorithm and symbol recognition technology to measure the position and attitude of UAV.The measurement error is calculated and the reasons inducing the error is analyzed.In addition,the application of GPU to achieve accelerated TLD algorithm is implimented and tested.First of all,the Zhang Zhengyou calibration method is used to calibrate the monocular camera,and the camera internal parameters and distortion parameters are obtained.It is the basis of position measurement.Secondly,based on CamShift algorithm,the position and attitude measurement of UAV target is realized.The static and dynamic model of the UAV are measured to calculate three target coordinates and three attitude angles.The range of the measurement precision are obtained,and the reasons are analyzed.Thirdly,to address the inaccuracy of CamShift based monocular vision measurement,the position and attitude measurement experiment is carried out by encoding the sign recognition algorithm.The measurement error is also calculated and compared with the results of thoes without sign recognition algorithm in the third chapter.Experiments show that the measurement error produced by the measurement method based on the recognition of coded symbols is restrained effectively and the accuracy of position and attitude measurement is improved.Finally,in order to improve the real-time performance of target measurement,the parallel computing advantage of GPU is used to parallelize the TLD tracking algorithm in CUDA architecture.The time consuming of each part of the TLD algorithm is computed and compared.The frame processing speed is increased from 16 frames per second to up to 22.4 frames per second,exhibiting great spedd-up cabability.