Research On Data Acquisition And Fusion Method Of Visual Inertial Odometry

Visual inertial odometry(VIO)is a new type of odometry,which uses both inertial navigation system and visual odometry technology.Through data fusion,it can output mileage data with higher accuracy than traditional odometry,so it is widely used in the field of unmanned driving technology and robot technology.The sensor data acquisition for visual inertial odometry needs to solve the problem of synchronization between visual data and inertial data.However,due to the unstable output frequency of visual data,the traditional software data acquisition method can not realize the synchronous and stable acquisition of visual data and inertial data.On the other hand,the existing research on visual inertial odometry calculation method usually uses inertial navigation data to optimize the front-end visual odometry motion estimation or to improve the output mileage accuracy of visual inertial odometry by using inertial navigation data,but it does not effectively and pertinently solve the problem of increasing mileage error when the carrier plane turns.Aiming at the above problems,this dissertation designs and proposes a data synchronization acquisition method for visual inertial odometry and an data fusion algorithm based on course angle error compensation.Firstly,this dissertation uses the method of multi process programming and time constraints of single group data acquisition to solve the synchronization problem of visual inertia data.Taking Dell laptop as the data acquisition platform,the hardware platform is built by using the MTi-G710 micro attitude reference system of Xsens company and AR0144 highdefinition binocular camera as the sensor.After data acquisition,it is not necessary to synchronize the offline data,which facilitates the research of visual inertial mileage calculation.Then,this dissertation proposes a data fusion algorithm based on the compensation of heading angle error,which improves the accuracy of plane turning mileage of the vision inertial odometry.In this dissertation,the error model of heading angle of inertia navigation system and binocular vision odometry is established based on the characteristics of the carrier making plane turning motion.The reason for the increase of the heading angle error of the two kinds of positioning algorithms is explained.The similarity of the change of the heading angle error between the inertia navigation system and the visual odometry is analyzed.The similarity of the heading angle error is used to analyze the similarity of the heading angle error between the inertia navigation system and the visual odometry The error compensation is made to the angle,which reduces the output mileage error of the visual inertial odometry.The validity of the above algorithm is verified by the result of the solution of the KITTI data set.Finally,this dissertation uses the experimental data collected by the self-designed and implemented data acquisition method to verify the effectiveness and error analysis of the proposed algorithm.The binocular camera is calibrated in the experiment,and the collected visual data is preprocessed.The experimental results show that the algorithm proposed in this dissertation can effectively reduce the mileage error of visual inertial odometry when the carrier plane turns.At the same time,the defects in the experiment are analyzed.