Design Of Autonomous Navigation System For Drone Based On Vision

This paper aims to design a vision-based drone navigation system for GPS-free environments.In view of this situation,This research has carried out a lot of research on the existing indoor UAV navigation methods.Based on this,a visual navigation algorithm based on RGB-D camera is proposed.In addition,in order to improve the accuracy of data,visual and inertia are also applied.Navigation is fused to optimize visual navigation data information.Apply this navigation algorithm to the indoor positioning of the drone.The main work of this paper includes:(1)In order to solve the problems existing in indoor vision algorithms,this paper proposes the following improvements to the RGB-D SLAM algorithm based on Xtion Pro live camera: 1.The ORB-based feature detection and descriptor extraction method is used for feature detection and Describe the sub-extraction phase and filter feature points where the depth information is not valid.2.Feature matching is performed using the KNN method used in the feature matching phase.3.The motion transform estimation method of ICP is used in the motion transform estimation stage.(2)The basic positioning principle of the inertial navigation system is studied,including the inertial navigation coordinate system,the basic rotation theory,and the attitude angle calculation for the drone.The attitude angle calculation based on complementary filtering is compared and the Kalman based expansion is based.The attitude angle solution.It laid the foundation for subsequent research and application.(3)In order to further improve the accuracy of indoor navigation data,a navigation algorithm based on inertial navigation and visual fusion is proposed.The IMU data is replaced by the initial posture in the visual SLAM.Through the complementarity of vision and inertial navigation,the accuracy of the pose estimation of the drone during the motion is improved.