Key Technologies Research On Stereo-visual/MIMU Integrated Navigation System

With the development of unmanned vehicle and intelligent robot,Machine Vision and Micro Inertial Measurement Unit based on MEMS Technology are widely used.Visual system is often used to perceive external environment and reconstruct three-Dimensional space,and then realize obstacle detection,object tracking and navigation.Positioning technology based on vision can provide accurate location information,but visual system is sensitive to illumination change.Therefore,when the light is insufficient or the sight is blocked,it will can not provide location information.Micro inertial navigation system has the characteristics of small size,light weight and low cost,however it can only provide a short time high precision position and attitude information,which often need the assistant of GPS or other positioning system.This paper researches a positioning technology based on stereo vision system,which can provide location and attitude information for vehicle,and this method often called Visual Odometry,and a combined navigation system is proposed,which is based on the advantages and disadvantages of the visual system and micro inertial system,and it will provide accurate location information when the satellite signal is shield or failure.Firstly,this paper derived the position,velocity and attitude updating algorithms of Strapdown Inertial Navigation System and the error equation.Then,this paper carried out the research on key technologies of visual localization based on binocular CCD camera,including high precision stereo camera calibration,binocular system feature matching,consecutive frame image feature matching and motion estimation of vehicle,and proposed feature matching method based on FLANN and RANSAC with multiple constraints.Finally,this paper established linear Kalman filtering model for the fusion of visual and inertial data,which use the positioning error as the observations,and then experiments based on KITTI Datasets and data from MTI-G-710/ZED system were performed to testify KITTI datasets and data from MTi-G-710/ZED system For a better understanding and analysis of those key technologies,and fully analyzing has done for further research.