Research On Position And Attitude Measurement Method Of Multi Rotor UAV Based On Vision

In recent years,unmanned aerial vehicle(UAV)technology has developed rapidly,and the UAV has been used in various fields.Navigation and positioning technology is the core technology of UAV,and it is crucial to the performance of UAV.Vision technology has good local information acquisition ability and strong anti-interference ability.It can be used as an auxiliary technology for navigation and positioning of UAV.Under the background of the application of multi rotor UAV in automatic operation and maintenance of photovoltaic power station,this thesis studies the measurement method of position and attitude of multi rotor UAV based on vision.The main research contents are as follows:Firstly,the overall design of vision based position and attitude measurement system for UAV is presented.The basic principle of UAV position and attitude measurement based on projection transformation is studied and the performance index of the system is determined.The dual circle target is selected as the reference target,and the world coordinate system is established.The coordinates of the 21 feature points in the world coordinate system are determined.The transformation relation from the world coordinate system to the camera coordinate system is deduced,and the homography transformation of the feature points from the reference target plane to the image plane is further deduced.By solving the homogeneous linear equations corresponding to the homography transform ation,the approximate position and attitude of the camera coordinate system related to the world coordinate system are obtained.Secondly,the method of extracting and labelling feature points from images captured by camera is studied.The image captured by the camera is pre-processed and converted into binary image.Then,the Two-pass connected component labeling algorithm is used to label the binary image,and the reference target is extracted from the labeled image.An edge detection algorithm and a curve fitting algorithm are used to obtain the equation of the ellipse curve of the reference target.Using the rolling tangent method proposed in this thesis,all 21 feature points are extracted and labeled.Finally,an optimization scheme for the approximate solution of posi tion and attitude of UAV is given,and nonlinear optimization is carried out by using Powell’s Dogleg algorithm.A experiment platform was built and 20 images were captured for experiment.The experimental results show that the average absolute value of single axis error were 17.95 cm,11.50 cm,3.65 cm and the average absolute value of single axis attitude error was 8.43 degrees,9.11 degrees,0.56 degrees.This indicates that the accuracy of the UAV position and attitude estimation is relatively precise.