Research On Positioning Technology Combining Binocular Vision And Inertial Measurement Unit

With the continuous development of technologies such as unmanned driving,artificial intelligence and mobile robots,visual inertial integrated navigation and positioning technology has gradually become one of the indispensable research directions in these research fields,and has received widespread attention from scientific researchers.The visual inertial integrated navigation and positioning system combines a low-cost,miniaturized inertial measurement unit(IMU)and a camera,which can not only use the rich scene information provided by the camera to correct the cumulative drift of the IMU,but also use the carrier output by the IMU itself The motion information provides more accurate initial pose information for the binocular vision system,and the carrier's position and attitude in the inertial space are solved by the fusion of the two information.The system makes full use of the complementary relationship between vision and inertia for the navigation of the carrier And positioning.In this paper,the motion state of the carrier is estimated by the navigation and positioning system composed of binocular vision camera and MEMS inertial measurement unit,and the nonlinear optimization and tight coupling scheme with higher navigation and positioning accuracy is used to solve the position and pose of multi-sensor information fusion.First,define the coordinate system involved in the binocular visual navigation system and clarify the transformation relationship from the camera coordinate system to the world coordinate system,analyze the camera's Zhang's calibration method,and comparatively analyze the four SIFT,Harris,ORB and SURF A mainstream image feature point detection method,which further verifies the optimality of the ORB feature point detection method,and based on the RANSAC method,a feature point removal algorithm for mismatching is given.The ICP method is used to estimate the motion state of the carrier.Extract the matching and track the image features and give a reconstruction algorithm based on ICP/SFM fusion;secondly,the coordinate system definition of MEMS inertial measurement unit and the conversion method between coordinate systems are given,and the measurement model of IMU is established.On the basis of pre-integration processing of IMU data,the posture information obtained by vision and inertia is processed by a loose coupling algorithm,and the accurate initial system value required for posture optimization is obtained by solving the linear least squares equation,including posture initialization,The gyro's bias,the direction of gravity acceleration and the motion state of the carrier;again,using the IMU measurement error and visual reprojection error to construct a nonlinear optimized cost function,based on the sliding window marginalization method to ensure that the visual navigation system constant Calculate consumption and real-time,at the same time,add marginal prior information to the cost function to avoid the loss of image information.In addition,a loop detection method based on the bag of words model is introduced to reduce the cumulative error of the visual navigation system and increase The four-degree-of-freedom position and pose optimization function ensures the global consistency of the integrated navigation system's motion trajectory,and provides a solution based on extended Kalman filtering for the problem of pure rotation degradation of the carrier;finally,the Euroc dataset was used for system simulation verification.