| High precision inertial navigation system can complete the navigation and positioning without external information source.Because of the inherent properties of its sensors and the inability to eliminate the accumulated error,it is necessary to introduce GPS for integrated navigation to improve the accuracy.However,the instability of GPS signal limits its universal applicability.Visual simultaneous location and mapping(VSLAM)technology can also complete the navigation and positioning without external information source.It collects the image information of the surrounding environment through the visual sensor,calculates the position and pose of the main body of the sensor by using the knowledge of multi view geometry,and then constructs the map and motion trajectory consistent with the surrounding environment.One of its major drawbacks is that it needs rich visual texture features,which is just the inertial navigation system The advantage of the system.To solve this problem,this paper is divided into three parts to explore.(1)The pinhole camera model and three-dimensional rigid body motion are briefly introduced.The internal parameters are obtained by camera calibration,and the points in three-dimensional space are projected to the two-dimensional image plane.The traditional fast feature points are improved,and the accuracy and rotation invariance are increased without changing the rapidity.In order to improve the accuracy of feature matching,the idea of RANSAC algorithm is integrated into FLANN feature matching,which not only reduces the number of feature points matching,but also retains the most accurate feature matching.In the visual initialization module,homography matrix and basic matrix are calculated in parallel,and different starting methods are selected according to different scenes.This paper briefly introduces the key frame selection strategy of vision slam and the construction process of motion trajectory,and runs the improved system in the actual environment.It is found that the system can not operate normally in the environment with missing visual texture.It is proposed to add inertial attitude measurement system as a supplement to the visual sensor.(2)The Kinect camera is added with an inertial attitude measurement system.The ROS system is used as the medium to receive inertial data.The stm32f103vet6 is used as the MCU,and the mpu-6050 is used as the inertial measurement unit to design an inertial attitude measurement system.The clock and reset circuit,power supply and download circuit,and the communication module with ROS are briefly introduced.Aiming at the noise problem of MEMS devices,iir filter is introduced as an improvement.Data are collected on the turntable,and the data curves before and after iir filter processing are drawn in MATLAB.The curve fitting and error of the system at different rates are calculated to verify the effectiveness of iir filter.The experimental results show that the system can effectively measure the position and attitude information of the carrier.(3)Aiming at the problem of how to deal with inertial data,the noise model and kinematics model of IMU are established firstly,then the discrete form of IMU pre integration is introduced for programming,and the initialization module of the system is improved to include inertial data.Finally,the inertial data and visual data are put into the same state equation by using the idea of tight coupling,and the iterative calculation error is minimized.The edge technology is introduced in the sliding window to keep only one key frame,which ensures the real-time performance of the system without affecting the accuracy of pose estimation.Thirdly,experiments are carried out in the environment of missing visual texture,and the results show that the system can operate normally,and the accuracy indicators are improved by 50% to 70%.The experimental results of tum data set show that even in the environment of rich visual texture,the accuracy indicators of the system can still be improved by 10% to 50%. |
PDF Full Text Request