Research On Positioning And Attitude Determination Algorithm Of Inertial/Visual Integrated Navigation Based On Optical Flow

With the rapid development of navigation technology,its application range has become more and more extensive,both in the civilian and military fields.Among them,inertial/visual integrated navigation has gradually become one of the important development directions in the navigation field due to its good complementarity and autonomy.Inertial navigation uses inertial devices to perceive the movement information of the carrier itself,does not need to introduce external information,and has high autonomy.However,its calculation error will accumulate over time.A large number of biological experiments have shown that insects such as bees and ants can use visual information to perform precise navigation tasks.When the insect moves in the environment,the image movement on the retina will generate an optical flow signal,which provides rich spatial feature information for the insect's visual navigation.In this context,this paper draws on the optical flow navigation mechanism of insects,uses visual information to assist the inertial navigation system,and studies the inertial/visual integrated navigation positioning and attitude algorithm based on optical flow.Firstly,the principle of inertial/visual integrated navigation based on optical flow is introduced,which includes three parts: common coordinate system of inertial/visual integrated navigation,strapdown inertial navigation principle and optical flow-based visual navigation principle.The principle of strapdown inertial navigation explains the attitude,speed and position update algorithm of strapdown inertial navigation,while the optical flow-based visual navigation part mainly focuses on optical flow,optical flow field and the solution of navigation parameters.Secondly,the LK(Lucas-Kanade)optical flow algorithm based on improved ORB(Oriented FAST and Rotated BRIEF)feature points is studied.On the basis of traditional ORB feature points,the Hessian matrix constructed by the SURF(Speeded Up Robust Features)algorithm in the feature extraction step is introduced,and an improved ORB feature point extraction algorithm based on the Hessian matrix is proposed.And we carry out feature matching performance experiment and feature tracking experiment based on LK optical flow method to test the performance of improved ORB feature points.The experimental results show that compared with the traditional ORB feature points,the improved ORB feature points have a significant improvement in feature point matching and LK optical flow tracking.Then the 3D motion parameter recovery algorithm based on optical flow is studied.Based on the Singular Value Decomposition(SVD)theory,two SVD-based homography matrix decomposition methods including Faugeras SVD decomposition method and Zhang SVD decomposition method are compared.Zhang SVD decomposition method is selected after comprehensive analysis.And we use the KITTI computer vision data set to verify the real-time performance and accuracy of the above algorithm.The experimental results show that the 3D motion parameter recovery algorithm based on optical flow has high accuracy,and it also has the possibility of further combination with inertial navigation system in real-time performance.Finally,based on particle filter technology,an inertial/visual integrated navigation model is established,which includes a complete state equation and measurement equation.We also use the KITTI computer vision data set to verify the accuracy of the above-mentioned integrated navigation system.The experimental results show that the particle filter-based inertial/visual integrated navigation system has higher accuracy than the Kalman filter,and can meet the navigation requirements of vehicle-mounted scenes.