Research On An Improved Positioning Algorithm Based On The Fusion Of Vision And Inertial Navigation

With the continuous improvement and intelligence of mobile robots,the simultaneous localization and mapping(SLAM)system based on multi-sensor fusion,especially based on visual SLAM,has become a key area of current research.However,because of some problems with vision hardware such as cameras,when the mobile robot moves quickly,some of the image texture is missing or part of the feature information is missing,which will cause problems in image information collection,especially for monocular cameras.Issues such as scale uncertainty.An inertial measurement unit(IMU)is a motion sensor that carries three-axis acceleration and angular velocity.Both can provide scale information and pose estimation information for the monocular camera,but the IMU alone has the problem of static drift.Therefore,the fusion of IMU and visual SLAM can simultaneously give play to the advantages of the camera and IMU sensor,and improve the accuracy and real-time performance of the SLAM system during the robot positioning process.This paper first carefully studies the existing SLAM system that integrates vision and IMU,and proposes an improved SLAM system that integrates inertial navigation vision.At the front end of the vision,the image is divided into blocks for the easy missing and bunching phenomenon of FAST feature points,and the FAST feature points in each square are extracted with a fixed threshold,and then Shi-Tomasi scoring is performed,and the extracted FAST feature points are lighted.Flow tracking.In the process of tracking,the random sampling consensus algorithm is used to eliminate the outlier points that appear.Secondly,in view of the problem of visual information and IMU information being out of sync,the IMU pre-integration method is used to solve the problem of inconsistency between the IMU frequency and the image frequency.In the process of fusion of vision and IMU,a tight coupling method is used to establish constraints between the pre-integrated residual term of the IMU and the reprojection of the camera,and a new damping update algorithm is used to reduce the calculation iteration time.,Improve the real-time performance of the system.In order to solve the problem of the accumulated error of the pose information during the movement of the mobile robot,a perceptual hash algorithm is proposed on the traditional loop detection algorithm,and the hash sequence of the previous frame and the next frame is compared to determine whether the loop is looped.Then optimize the system globally to eliminate accumulated errors.Finally,the algorithm designed in this paper is verified on the Euroc standard data set.The experimental results show that this paper can extract evenly distributed feature points and the tracking effect is good,the robot positioning accuracy error is small,and the whole system has good real-time performance.