Research On Optimization Method Of Vision/Inertial Fusion Navigation

In modern navigation system,integrated navigation is widely used because of its relatively accurate pose calculation and stronger robustness.With the rapid development of mobile robots,drones and other intelligent devices,SLAM(Simultaneous Localization and Mapping)technology using laser,vision,INS,GNSS and other sensors has also attracted more and more attention and research from scholars.In integrated navigation,when faced with the lack of external information such as GNSS,visual measurement can use image information to continuously correct the zero offset and zero drift error of inertial measurement.Inertial measurement also provides reliable auxiliary information for visual measurement.The use of vision and inertial devices for integrated navigation has significant advantages.However,although scholars have continuously optimized the error and system robustness in the algorithm framework of vision/inertial navigation system,there are still shortcomings of insufficient error optimization and poor adaptability to real scenes.Aiming at the shortcomings of the vision/inertial system,this paper focuses on the error optimization method and the map incremental optimization model.The main contents are as follows:(1)Propose stochastic model optimization method.Due to the imaging characteristics of the vision sensor,there are random errors in the measurement process.By clustering the visual measurement information according to the reliability,optimize the estimation of variance-covariance components,establish a scientific and reliable optimized communication form to improve the accuracy of the posture calculation of the visual/inertial fusion navigation algorithm.(2)Propose a map incremental optimization model.Construct a reusable visual inertial map in an indoor environment and realize incremental expansion of the map,establish adaptive processing front-end processing to achieve long-term navigation and positioning needs,score the reliable information on the back-end map processing part and propose a constrained optimization equation.Experiments have shown that the incremental map optimization model can reduce the impact of gross errors and system errors in the long-term navigation and positioning system.The system has a certain ability to adapt to environmental changes and reduces the memory usage of map information to enhance the robustness of the system.(3)Establish a navigation system based on vision/inertial fusion navigation model.Based on the vision/inertial fusion navigation algorithm,the system adds a random error optimization module and a map incremental optimization module,and uses a self-built car equipped with a binocular vision/inertial fusion navigation algorithm for practical application and experimental analysis.