An Autonomous Positioning System For Quadrotor UAV Based On Vision And IMU

At present,the positioning method of machine vision has the advantages of high positioning accuracy,rich information collection,low cost and wide application range,which has become the mainstream trend of research and application of unmanned aerial vehicle autonomous positioning system.The commonly used Simultaneous Localization And Mapping(SLAM)technology is based on the Simultaneous Localization And Mapping(Simultaneity)technology to estimate the position and attitude of the map key frames and feature points collected by the airborne camera,which is easily affected by the flight state of the UAV,the performance limitation of the airborne image processing module,environmental conditions and other factors.Aiming at the problems related to map information collection and processing in the autonomous navigation system of quadrotor UAV based on visual SLAM,an autonomous positioning system of quadrotor UAV based on visual and inertial sensor(IMU)is designed and implemented in this thesis.First of all,to solve the problem of the short lack of map information,a four-rotor UAV positioning system based on vision and IMU is designed,and the position information obtained from IMU is used as a supplement to visual SLAM.The hardware design and software design of the autonomous positioning system of quadrotor UAV based on vision and IMU are introduced in detail.In terms of visual and IMU information processing,a tightness coupling nonlinear optimization model based on sliding window is designed to improve the positioning accuracy of the system,and the calibration of the visual and IMU sensors is completed,thus reducing the influence of sensor measurement noise on the positioning accuracy of the system.Then,an improved Retinex image enhancement algorithm is designed to improve the quantity and quality of feature point extraction from UAV images under uneven or low illumination.In order to reduce the calculation amount of Retinex algorithm,the RGB images are transferred to HSV space,and the high-frequency images with rich edge features are intercepted by the homomorphic high-high-pass filtering algorithm for the V layer.For edge feature enhancement,the bilateral filter is combined with Retinex algorithm to reduce the problem of edge feature ambiguity.Simulation experiments verify the effectiveness of the improved Retinex algorithm.Secondly,an improved ORB feature matching algorithm based on sparse optical flow method is designed in order to meet the real-time,robustness and accuracy requirements of the image feature matching algorithm for UAV under the changeable environment.KMENAS ++ algorithm,sparse optical flow method and ORB algorithm are combined to eliminate the feature matching point pairs far away from the clustering center,and then the final feature matching result is optimized by the random sampling consistency algorithm.The simulation results show that the improved ORB feature matching algorithm can be robust,real-time and improve the algorithm accuracy.Finally,In order to verify the positioning system designed in this thesis,tests are carried out in public data sets and in real scenes.The feasibility of the system is verified on the open data set,and then the actual measurement experiments such as running trajectory and position information are completed in different scenarios.The experimental results show the feasibility and effectiveness of the system in this thesis.