Research On Ultra-close Distance Pose Measurement Technology Of Non-cooperative Spacecraft Based On Point Cloud

In recent years,the applications based on the technology of spatial non-cooperative target capture,such as spacecraft on-orbit maintenance,abandoned satellites' clean up,space attack and defense and automatic rendezvous and docking,have become a new development direction in the field of space on-orbit service.Non-cooperative target relative position and attitude measurement is the key to the final stage of non-cooperative target capture technology.In order to provide a technical reference for solving the problem of the relative pose measurement of close non-cooperative targets,a research has been made on the relative pose measurement over a close distance of non-cooperative spacecraft based on point cloud data in this dissertation.Firstly,this paper presents a method of relative pose estimation based on object model and global matching,aiming at the situation that the object model has been known and feature matching method has low precision and poor stability.This method focuses on the global automatic matching problem of the camera point cloud in the observation coordinate system and the target model point cloud.Get the target data point cloud with the depth camera,estimate the translation field between the data point cloud and model,structure solution space.Search the global registration result in the solution space with global ICP algorithm,obtain the target initial pose.In the sequence frames,get continuous pose result with the improved ICP algorithm based on point cloud edge and RANSAC algorithm between the transformed camera point cloud and models.The simulation and ground test results show that the algorithm has high precision and good robustness.Secondly,a method based on natural features of the 2D/3D data fusion is proposed,aiming at the situation that the object model is unknown.This method focuses on how to use the depth camera amplitude image for aiding the natural features detection of target data point cloud,and establish target coordinate system for pose estimation.Detect corner respectively from camera amplitude image and point cloud using 2D harris and 3D harris corner detection algorithms.Then get several feature points with bidirectional matching between 2D/3D corners.According to these feature points,establish the target coordinate system.With the feature points coordinates in the target and camera coordinate system,original pose is calculated.In the sequence frames,use improved ICP algorithm of this article between previous and current frame for continuous pose measurement.The simulation results show that the algorithm is feasible,but has cumulate errors.In addition,a software of pose measurement is developed.The software has complete function include open/close camera,get/save/display the image,parameter settings,point cloud filtering,point cloud registration,display of results and communication with the server.A exerimental platform has been set up based on the MESA SR4000 TOF camera,ground experiment has been conducted to verify the method of pose estimation based on object model and global matching and the improved ICP algorithm.The experimental results show that this method has high accuracy and good stability,meets the requirements of the project.