Study On Visual Inertial Navigation Algorithm Under The Constraint Of Structural Characteristics

As visual navigation system and inertial navigation system has good performance on complementation and automation,visual-inertial navigation system has gradually become a new research hotspot in the field of navigation.At present,there are two mainstream methods in the Visual-Inertial Navigation algorithm,which are the combination form based on filtering and the combination form based on optimization.For the combination form based on of filtering,only the current positioning information can be corrected,and the historical trajectory can't be corrected.With regard to the combination form based on the optimization,the whole trajectory can be optimized and error compensated by constructing the structure of the graph.However,as time goes on,more and more data are needed to be processed in the optimization algorithm,and the speed of computation will be affected.In order to meet the requirements of real-time performance in practical applications,the following strategies are used to improve the speed of computation in this paper:For the data of the visual part,Keyframe is firstly introduced,and then the key frame is selected before the change of the position and direction of the mobile platform by the optimization algorithm,so as to reduce the number of image processing.Secondly,introduces the concept of sliding window,through the edge technology some state variables are selectively removed from the estimator,thus data need to be optimized is bounded in a window,and the whole pose is optimized only when the closed loop is detected.The optimization variables only include the pose parameters of the carrier,but not the spatial position parameters of the feature points.For IMU data,the sampling frequency of IMU is much higher than the sampling frequency of the camera,so if we directly add the IMU data to the optimization algorithm,the cost of calculation will be greatly increased.The method of pre integration is used in this paper.First the IMU data between two key frames are pre integrated.Then the visual part is combined and optimized.The method can greatly reduce the amount of data being optimized,improving the calculation speed and ensure the real-time performance of the system.