Long-distance Relative Positioning Combined With Vehicle Sensing Data

The precise positioning of the lunar patrol is the premise and basis of the safe movement and path planning of the patrol,and also the basic guarantee for ensuring that the patrol is gradually approaching the scientific goal of the long distance and completing the scientific exploration task.Since there is no satellite navigation system on the lunar surface,the relative navigation is mainly used to achieve lunar surface positioning.Considering the particularity of the lunar surface environment,it is difficult to meet the positioning requirements with a single relative positioning method at present: 1)due to the many bumps and slopes of the lunar surface,the inertial navigation positioning will generate a large cumulative error,which is difficult to meet the requirements of long-distance positioning;2)due to the single texture of the lunar surface and complex lighting conditions,image feature mismatch is very likely to occur during visual positioning,and local positioning deviation is likely to occur.On the basis of analyzing the advantages and disadvantages of various relative positioning methods,this paper combines multiple relative positioning methods,and fully exploits the advantages of various positioning methods through multi-sensor fusion algorithm to improve the accuracy and reliability of long-distance navigation and positioning of the patrol.Sexuality solves the problem of long-distance relative autonomous positioning of the analog patrol.The specific research contents are as follows:(1)An image matching method based on improved ORB(Oriented FAST and Rotated BRIEF)features and structure constraints was designed to improve the matching performance.In this paper,three main feature extraction methods are analyzed,and the improved ORB feature extraction operator is proposed based on their advantages and disadvantages.Then,the images between adjacent frames are matched by graph matching method,and the motion trajectories between cameras are estimated by using the feature points well matched.Finally,the sequence images of the surrounding environment taken by the lunar patrol simulator were used to verify the improved ORB feature extraction algorithm and the graph matching method based on feature structure constraints,proving the effectiveness of the algorithm.(2)The ORB-SLAM(Oriented FAST and Rotated BRIEF-Simultaneous Localization and Mapping,ORB-SLAM)visual localization framework and its application in feature matching,3d mapping and continuous localization were studied,which laid the theoretical foundation for the improvement of multi-sensor fusion localization algorithm.In view of the continuous moving patrol unit positioning problem,this paper studies the camera imaging parameter model,the motion estimation method based on adjacent frame feature matching and positioning adjustment figure optimization based on the beam method,calculating model,focusing on high precision position based on nonlinear optimization calculating chart optimization and closed-loop detection methods are analyzed and the experiment.(3)The positioning technology system based on multi-sensor combination is studied,and the positioning technology method based on inertial navigation and binocular camera fusion is proposed to realize long-distance real-time navigation positioning.Aiming at the problem of large deviation error and error prone in the positioning of a single sensor,the navigation and positioning algorithm based on the IMU(Inertial Measurement Unit)and binocular vision fusion of the orb-slam framework was designed,and the mathematical models of inertial navigation and binocular vision positioning as well as their combined solution methods were given,and the combined solution algorithm was designed.The practicability and effectiveness of the algorithm in complex environments are verified by experiments,and the results show that the localization accuracy is better than the visual SLAM algorithm.