3D Reconstruction Method Of Non-cooperative Space Targets Based On Parallel Computing

For the needs of space development and orbital utilization,the focus of the research on space on-orbit service technology has shifted to the maintenance and disposal of the high-orbiting satellite.Obtaining the three-dimensional structure model of the target is an essential step in the orbit service task.However,the proportion of The space non-cooperative target is getting more and more in on-orbit service targets.The method based on the three-dimensional structure model to judge the target function model to capture object,docking tag,maintenance parts has great significance to space non-cooperation target arrest,drag and maintenance.Therefore,the acquisition of the three-dimensional structure of the space non-cooperative target is the inevitable demand to the smooth implementation of space on-orbit service.At present,there are a lot of researches on the three-dimensional reconstruction technology to obtain the three-dimensional structure model of the target.However,it is more applicable to the ground scene as the background of the target.In view of the particularity of spatial background and non-cooperative targets,the three-dimensional reconstruction technology of space non-cooperation target has its different research difficulties from other researches.At the same time,under the condition that the load of the computer is limited,the three-dimensional reconstruction technology of the space non-cooperation target must have the processing capacity of massive data and the corresponding reconstruction efficiency.In addition,the 3D reconstruction accuracy is also a key issue in spatial relative navigation.However,for the space non-cooperation target,the complete three-dimensional reconstruction accuracy evaluation system is rare.At the same time,whether the three-dimensional reconstruction system to meet the results of the reconstruction of the indicators isthe important basis to determine whether the system to meet the project delivery requirements.The main contents of this paper are:Firstly,the camera model and frame definition of three-dimensional reconstruction technology are described.Based on the image obtained by CCD camera,this paper expounds the working mechanism of the SIFT algorithm and its advantages.The theoretical foundation of three-dimensional reconstruction is also emphasized.Secondly,this paper studies and realizes a three-dimensional reconstruction technique based on stereo vision in allusion to spatial background and target characteristics.At the same time,a sparse three-dimensional point cloud reconstruction method based on bundle adjustment is proposed.In addition,the method of dense three-dimensional point cloud restoration is described,and a lot of simulation experiments are carried out.Thirdly,in order to improve the capacity of three-dimensional reconstruction and the efficiency of reconstruction,the three-dimensional reconstruction optimization method based on heterogeneous parallel computing platform is studied.The principle of heterogeneous parallel computing,computing platform model and programming method are detailed.In addition,the code parallelization and simulation experiments are carried out on the functional module of the three-dimensional reconstruction of space non-cooperative target.The reasons for the possible results are analyzed.Finally,this paper studies the evaluation system of three-dimensional reconstruction of space non-cooperative target,and introduces the basic principle of evaluation system.In addition,a lot of experiments based on a large number of simulation images have been carried out to verify the integrity and accuracy of 3d reconstruction technology introduced in this paper.