Algorithm Research On GNSS/SINS/Vision Integrated Navigation In Dense Urban Areas

At present,the development of unmanned moving platform technology in intelligent transportation field puts forward higher requirements for the accuracy and reliability of autonomous navigation.Satellite navigation,inertial navigation and computer vision navigation are the most widely used navigation methods in this field.However,the single navigation method has its inherent defects:In the dense urban environment of many skyscrapers,the radio signal of satellite navigation system is vulnerable to accidental or intentional occlusion and it will be impacted by incidental or intentional electromagnetic interference while it has the advantages of high positioning accuracy and the error does not accumulate with time.Inertial navigation can keep the precision and stability of short time navigation without relying on external information,but its error accumulates with time.Dense urban environmental has rich image features,but the image information is large,the calculation processing is complex,the real-time implementation is difficult and it is affected by lighting and other external factors.Therefore,the integrated navigation method of multi-sensor data fusion has become the focus of research in the related fields at home and abroad.This paper firstly analyze the performance of different GNSS/SINS integrated navigation mode and filtering algorithm,and then draw into the visual information and extract navigation,position and attitude constraints between images to aid inertial navigation system.Fusing the data of multi-sensor information by further analyzing the complementary among various navigation methods to maintain the navigation accuracy and stability while the satellite signals are blocked.When the satellite signal is restored,the attitude,velocity and position errors of the integrated navigation system will converge quickly,which can improve the adaptability of the system in dense urban areas and improve the navigation performance.The following are the innovations of this paper:(1)According to the actual running car data,the GNSS/SINS loose coupled and tightly coupled navigation algorithm are used to solve the navigation problem.In the navigation process,and the linear Kalman filtering and nonlinear Kalman filtering algorithm will be used to verify the navigation accuracy with short time of interference or shelter of the satellite signal.(2)Introducing visual navigation system based on the GNSS/SINS integrated navigation system according to the basic principle of visual odometry for multi sensor data fusion.The scene depth information is obtained by the stereo vision system,and the multi view geometric constraint Kalman filtering algorithm is used to perform VINS integrated navigation and design the filter.(3)Using the model errors of each navigation system as the state variables in the GNSS/SINS/Vision navigation system.Design the unscented Kalman filter and deduce the observation model of the integrated system.To verify the navigation accuracy by experimental data and compare with the two navigation methods above in the absence of satellite signals,which provides the basis for further improvement of the integrated navigation method.