Spaceborne Optical Image Based Space Target Pose Estimation Technology

The on-orbit status of spacecraft,especially the three-dimensional pose estimation,is an important basis for many aerospace missions,such as rendezvous,docking with space targets,maintenance of faulty satellites and so on.Compared with the two detection methods of ground-based radar and ground-based optics,space-based optical imaging has superiority in higher resolution,free from atmospheric turbulence and border restrictions,which can provide solid support for precise three-dimensional pose estimation of space targets directly.According to the difference of orbital motion characteristics of the observed objects,this thesis considers two kinds of space target attitude estimation problems based on space-based optical imaging,and they can be summed as pose estimation problem in the fast rendezvous scene of low-orbit(LEO)space targets and the pose estimation problem in the close orbit scenario for geostationary earth orbit(GEO)space targets.Focusing on the problems of pose estimation at the above two typical scenarios,this thesis explores the corresponding target attitude estimation methods as follows.When the space-based imaging platform observes the LEO target,the imaging observation view is limited by the rendezvous time and imaging distance,the effective numbers of target images pixels are small,and the total numbers of available observation images are limited.To address the problem of spatial target attitude estimation under such imaging conditions,this thesis proposes a LEO space target pose estimation method based on multi-model information fusion.Fusion effectively utilizes the complementarity of cross-model data in target representation and improves the accuracy of pose estimation results.The simulation results show that the method proposed in this thesis can accurately estimate the attitude angle of the LEO target when the feature information of the observed image is small and inconspicuous.Besides,it shows better adaptation in the degraded conditions such as noise and motion blur.When the space-based imaging platform observes the GEO target,the imaging process is often accompanied with the uneven illumination.And the adjustment of the local structure angle often occurs in the GEO target for practical demands,therefore there is a certain local difference between the actual structure of the target and the prior three-dimensional model.Aiming at the problem of spatial target pose estimation under such imaging conditions,this thesis proposes a GEO target pose estimation method based on semantic key points.The feature association between the target and the known3 D model is used to solve the target pose.Simulation experiments show that the method proposed in this thesis can effectively improve the pose estimation accuracy of spacebased when the observed image is degraded.In addition,an optimization method of target poses and structure parameters based on the principle of minimizing the reprojection error is proposed,which can still estimate the reliable pose when the target has structural changes due to faults or demands of practice work.