Research On Positioning Technology Based On Monocular Vision Fusion Inertial Navigation

The spatial fusion technology of visual fusion INS can provide accurate spatial position information for the robot in autonomous navigation and positioning.The visual information and INS information can complement each other in the system,and the visual information can effectively correct the drift of the INS.At the same time,in the environment lacking visual information,the inertial navigation can provide the system with short and effective location information.The complementary information can effectively improve the robustness and accuracy of the system,but it also poses a huge challenge in data fusion.In this paper,the key techniques and difficult problems of visual fusion of inertial navigation information in robot positioning system are studied.A set of positioning system with monocular vision fusion inertial navigation with precision and robustness is designed.The system is experimentally verified in the environment.The main research work of this paper is as follows:(1)The problem of online self-calibration of camera and IMU external reference was studied.Firstly,the camera pose motion estimation method based on feature method and optical flow method is adopted in visual information processing.The feature extraction method based on region of interest is proposed to effectively solve the feature repeat extraction and feature distribution unevenness in the tracking process.Secondly,the IMU is modeled and the IMU pre-integration under the median integral is deduced.Considering the noise and drift of the system,the first-order Taylor expansion approximation method is used to effectively solve the IMU's repeated integration due to the IMU bias update in the optimization process Finally,according to the hand-eye calibration method,the over-determination equation is constructed by using the motion estimation of the camera pose and the result of IMU pre-integration.The SVD singular value decomposition method is used to solve the external parameter rotation matrix of the camera and IMU,and the validity and feasibility of the algorithm is verified on the common dataset.(2)Based on the on-line self-calibration of camera and IMU external reference,the initialization and back-end optimization of the positioning system of monocular vision fusion IMU is studied.Firstly,the system is initialized according to the scaled rotation matrix of the camera IMU,and the initial value required for the system operation is calculated.Then,based on the initial value,the optimization equation is constructed by the method of tight coupling between visual and IMU data,and the sensor time deviation optimization is added to it,which effectively reduces the local drift caused by the different sampling time of the sensor.Finally,the overall drift problem of the system is adopted by loopback detection and global pose optimization strategy to ensure the global consistency of the system,and the accuracy and robustness of the proposed positioning system are verified by experiments on the common dataset.(3)Experiment the verification of the positioning system of this paper through the experimental platform.The experimental platform of this paper consists of a differentially controlled Turtle Bot mobile robot and a camera module integrated with mpu6050.First,an internal reference calibration experiment is performed on the monocular pinhole camera to ensure accurate feature position during camera motion estimation.Then the experimental verification of the positioning system of this paper is carried out in different real scenes.