Vision Aided Navigation Algorithm

A single navigation system has its shortcomings,and the integrated navigation system can make up for the shortcomings.The development of vision technology in recent years has once again promoted the advancement of navigation technology.The use of vision-assisted inertial navigation system can achieve precise navigation,especially suitable for GPS failures in complex environments.At the same time,the use of visual assisted positioning system has important application value for military tasks such as formation flying and aerial refueling.Based on the above research background,this thesis takes the research of vision-assisted navigation algorithm as the core,and carries out the research of monocular vision/IMU integrated navigation algorithm,binocular vision relative navigation positioning algorithm and improved algorithm,making the navigation system more robust,Concealment and accuracy.The main research contents include:(1)Aiming at the problem that the navigation error of inertial navigation accumulates over time,an integrated navigation algorithm that uses monocular vision to assist inertial measurement unit(IMU)navigation is studied.Solve the navigation results using monocular vision and IMU respectively.Align the navigation results calculated by vision and IMU.Using the measurement residuals of vision and IMU,an objective function for nonlinear optimization is constructed.By minimizing the objective function,the navigation state is estimated.Experiments have verified the effectiveness of the algorithm,and the frequency and accuracy of the algorithm’s output navigation results can meet the navigation requirements of general moving bodies.(2)Aiming at the scale invariance of binocular vision,a relative navigation positioning algorithm for binocular vision based on ORB(Oriented Fast and Rotated Brief)algorithm is proposed.The positioning algorithm uses the ORB algorithm to track the feature points.Based on the tracked feature points,a relative navigation positioning scheme is designed,and a binocular vision relative navigation positioning model is established.With the UAV formation as the experimental background,the relative navigation and positioning experiment is carried out.The experiment verified the effectiveness of the established binocular vision relative navigation and positioning model,and also found that the ORB algorithm has the problems of poor accuracy,large amount of calculation,and low real-time performance in practical applications.(3)Aiming at the problems of poor accuracy,large calculation amount,and low real-time performance of the ORB algorithm during UAV formation flying,the ORB algorithm is improved,and an improved binocular vision relative navigation and positioning algorithm is proposed.Improved ORB algorithm reduces the calculation amount of ORB algorithm,and also improves the tracking quality of feature points.The UAV feature points extracted by the improved ORB algorithm are substituted into the binocular visual relative navigation and positioning model,and the relative navigation and positioning between the formation UAVs is realized.The Kalman filter algorithm is used to estimate the relative navigation and positioning information of the UAV,which further improves the positioning accuracy.Experiments show that the improved binocular vision relative navigation and positioning algorithm has high accuracy and real-time performance,and has engineering application value.