Research On Relative Navigation Method Of Space Uncooperative Target

On account of the non-a priori characteristics of the target shape and relative motion information in the relative navigation mission of uncooperative spacecraft,and considering the requirements of robustness,real-time and autonomy in the space relative navigation process,this paper approaches the uncooperative spacecraft rendezvous.The research on relative navigation methods mainly focuses on the extraction and matching of image features at front end,target point cloud registration and splicing,target point cloud reconstruction and optimization at back end,and relative navigation information optimization.Main contributions of this thesis are as follows:In this thesis,the geometric projection principle of monocular camera and parallel binocular camera are analyzed.the imaging models of monocular and parallel binocular cameras are established respectively,and the influencing factors of camera measurement accuracy are analyzed.The principle of lens distortion is analyzed and a camera distortion correction model is established.The calibration principle of the binocular camera is analyzed.The three-dimensional software is used to generate the checkerboard image sequence input calibration algorithm to complete the camera calibration simulation experiment,and the precise internal parameters of the camera are obtained.The parallel strategy of sparse reconstruction and dense reconstruction not only meets the real-time requirements of navigation information update,but also ensures the accuracy of three-dimensional reconstruction of the target.For the sparse reconstruction process,the influence of space environment and close proximity on the imaging of camera is analyzed.The quantitative,accurate,real-time and invariance of ORB,SIFT and SURF algorithm are compared through simulation analysis.The analysis focuses on the invariance of rotation,scale,blur and illumination of each algorithm,and selects the feature extraction algorithm that satisfies the robustness requirements of sapce uncooperative target relative navigation mission.The feature points are realized by cross-check and MLESAC exact matching algorithm,and reconstruct a single frame sparse point cloud of the target by using a binocular imaging model.For the dense reconstruction process,the target disparity map is generated by the polar line search and block matching algorithm,and the single-frame dense point cloud of the target is reconstructed by the imaging model.The relative pose model of the spacecraft is established,and the influencing factors of the accuracy of the iterative closest point algorithm are simulated and analyzed.This algorithm is used to solve the point cloud pose transformation sequence of the motion process,and the point cloud is spliced frame by frame to obtain the target point cloud.In addition,the point cloud filter is used to filter the original point cloud to obtain a high-quality point cloud,and the optimized target point cloud is meshed and transformed into an obstacle constraint model.Finally,the geometric relationship between the relative motion of the spacecraft and the point cloud pose is analyzed,and the relative navigation information of the spacecraft is restored from the point cloud pose transformation sequence.A complete framework of space uncooperative target relative navigation algorithm is established,and a complete simulation experiment of navigation algorithm is carried out.Lie group and Lie algebra are introduced,and the Bundle Adjustment method based on graph optimization is used to optimize pose information and improve navigation accuracy.