Research On Visual/INS/GNSS Multi-Sensors Fusion Vehicle Positioning And Orientation Algorithm

In the urban environment,the vehicle Positioning and Orientation System(POS)mainly relies on the combination of GNSS and INS,but this method is easily affected by the strength of GNSS signals,especially in high-rise buildings,alley and other more serious scenarios which GNSS signal are blocked or missing.If the visual sensor or other on-board sensor can be used to assist the INS Positioning and Orientation,then the complementary characteristics of the multi-sensors fusion can be utilized,then the deficiency of the traditional vehicle Positioning and Orientation determination system can be effectively compensated.The research on vehicle GNSS/INS combined Positioning and Orientation system on domestic is relatively mature,but the Visual/INS/GNSS integrated navigation system is less studied.This thesis studies the key technologies of multi-sensors data fusion such as Visual Sensor,IMU,GNSS and also on.Meanwhile we applies it to the actual urban environment for Land Vehicle Positioning and Orientation.And we also use both simulation data and vehicle-borne measured data to verify our proposal algorithm.The main research contents are as follows:1.ECEF is used as the INS navigation reference coordinate system to design and implement the high-precision inertial navigation mechanical programming algorithm.The analysis shows that our algorithm formula is more concise and easier to implement.Meanwhile the INS can be better coupled with GNSS Position,Velocity,Pseudorange,Carrier phase and Doppler observations.2.Based on the Kalman filter theory,a multi-sensors fusion mathematical analysis model for the Visual/INS/GNSS integrated navigation system is established.The sensors such as GNSS,vision sensor and odometer are coupled with the INS navigation system.It is easier to implement and expend because of its clear structure,and its advantage for expansion and transplantation.3.INS/GNSS integrated navigation system could help vision sensor to improve image feature point matching accuracy;INS/GNSS integrated navigation system can provide higher accuracy Positioning and Orientation information in short time,which can be used for prediction of essential matrix,fundamental matrix and 3D coordinates of feature points etc.The experiments show that the method can effectively improve the accuracy of feature points matching.4.Visual/INS/GNSS multi-sensors fusion algorithm are developed on Linux platform.And the fusion algorithm is based on the matrix operation library Lapack/Eigen.And the open source computer vision algorithm library OpenCV is used to process the extraction and matching of image feature points.Both simulation data and Land Vehicle measured data are used to test our fusion algorithm in several times.The results from experiment show that our proposal algorithm can be run well.5.Visual/INS/GNSS multi-sensor fusion algorithm verification analysis.The Experiment which using both simulation data and actual vehicle-borne measured data shows that the Visual/GNSS/INS multi-sensors fusion algorithm implemented in this thesis can effectively increase the position and orientation accuracy by 40%?60% approximately.The relative position and orientation measurement data output by the visual sensor can also effectively improve the observability of INS/GNSS integrated navigation system.