Research On Mimu/Monocular Vision/Geomagnetic Integrated Positioning And Orientation Algorithm

In recent years,the problem of multi-sensor positioning and orientation without satellite signal has become a research hotspot.With the development of micro--mechanical technology,while the performance of MEMS-IMU is improved,the volume and cost are also continuously reduced,so it has been widely used.Magnetic sensors can effectively determine the heading angle according to the characteristics of the geomagnetic field,but there are defects that are susceptible to electromagnetic interference.Monocular cameras can obtain rich image information,but there are defects of uncertain scale.In this project,MEMS-IMU,magnetic sensor and monocular camera are used to design the integrated navigation system,which can realize the navigation task in indoor and outdoor environment without satellite signal.The main contents of this paper are as follows:The coordinate system involved in the system is clearly defined,the coordinate system used in this project and three attitude representation methods are introduced,and the advantages and disadvantages are analyzed;The measurement model of the IMU and the pinhole model of the monocular camera are established;the overall scheme of navigation system design is given.Based on the gradient descent method,an attitude fusion algorithm of IMU and magnetic sensor is designed,which effectively solves the problem of low accuracy and divergence of long-term work when using the gyro alone to update the attitude;Aiming at the influence of motion acceleration on the accuracy of attitude fusion algorithm,two adaptive algorithms are designed.The integrated navigation algorithm is designed based on the method of sliding window optimization.At the front end of the system,Harris corner detection + LK optical flow method is used for image processing;Use pure visual SFM for visual initialization,pre-integration for IMU initialization,and loose coupling of visual and IMU information to achieve joint initialization;At the back end of the combined system,the objective optimization function is constructed by IMU residual,visual residual and prior information,and the nonlinear optimization is carried out by using the least square method.In order to improve the accuracy of the solution,the relocation and pose optimization modules are added to the system to ensure the global consistency of the system motion.The hardware platform and software system of algorithm simulation are introduced,and a physical test is designed to test the performance of the attitude fusion algorithm;The data of the Eu Roc dataset is used to verify the combined navigation algorithm in this paper,and the results are analyzed.Compared with other mainstream algorithms,it is shown that under the condition of turning on the loopback function,this algorithm can effectively reduce the accumulation of errors compared with other algorithms in long-term work,improve the robustness of the algorithm,and realize the function of vector motion trajectory estimation more accurately.