On-orbit State Estimation Methods Of Space Targets Utilizing Multi-view ISAR Images

On-orbit state information of space targets,such as the three-dimensional(3-D)structure and component orientation,is of great value for satellite load analysis and health status assessment,etc.And it is the focus of space situational awareness.Benefiting from the allday,all-weather,long-range,and high-resolution imaging capability,ground-based Inverse Synthetic Aperture Radar(ISAR)is widely applied in space situational awareness.However,ISAR imaging results can only reflect the two-dimensional projection information of the target on the Range Doppler(RD)imaging plane.The on-orbit state is implied in the projection information and cannot be obtained directly.Therefore,the research on on-orbit state information estimation for space targets utilizing ISAR images will effectively improve the information acquisition ability of ground-based ISAR.And conclusively,it will support the completion of various space situational awareness tasks.At present,there are some problems with the existing on-orbit state estimation methods utilizing ISAR images,such as insufficient robustness of wide bandwidth ISAR imaging,inaccurate imaging geometry analysis,insufficient consideration of imaging characteristics,and incomplete observation scene modeling.These problems lead to limited accuracy and applicability of existing methods.Aiming at the application requirements and research status,this paper studies key issues of on-orbit state estimation from ISAR images.Relevant work is funded by the National Natural Science Foundation of China,and the Joint Fund of the Ministry of Education.The research context is progressive.Firstly,the wide bandwidth ISAR imaging method is studied to provide high-resolution images for the space target onorbit state estimation.Then,under the scene that a stationary moving target is observed by single-station ISAR,we accurately analyze the imaging geometry,and establish the mapping relationship between the ISAR image sequence and the target state,including threedimensional structure and attitude of the target.Combined with the ISAR imaging characteristics,we effectively exploit the image sequence features to reconstruct the 3-D distribution of target scatterers.And with the 3-D structure priori,attitude parameters of the target are estimated.Further,the state estimation scenario is extended to complex moving space targets.Under the condition of unknown target motion and geometric model,the onorbit state estimation utilizing multi-station joint observation is studied to break through the priori constraints and improve the applicability of existing methods.The specific research content of this paper can be summarized into the following four parts:1.Wide bandwidth ISAR imaging and cross-range scaling for space targets With the increasing signal bandwidth of the ISAR system,the nonlinear range cell migration(RCM)caused by target equivalent rotation is significantly enhanced.It makes the image defocus easily.Aiming at this problem,this paper firstly analyzes the effect of target equivalent rotation on the coherence among wide bandwidth echoes.And based on the imaging principle of Polar Format Algorithm(PFA),the PFA imaging error with mismatched equivalent rotation parameter is deduced.Then we prove the feasibility of measuring the matching degree of PFA parameters by image entropy.After that,a target rotation parameter estimation and PFA imaging method based on minimizing image entropy is proposed.It simultaneously compensates the envelope and phase errors introduced by RCM,thus enabling focused imaging.And cross-range scaling for the fused image can be easily implemented with estimated target rotation parameters.Finally,the effectiveness of the proposed method is verified by simulation experiments.2.Trajectory association and 3-D reconstruction of scatterersThe accurate scatterer trajectory association is the basis of the target 3-D reconstruction.However,the electromagnetic scattering of scatterers is significantly anisotropic and mutual occlusion,which makes the trajectory association difficult.Aiming at this problem,this paper firstly derives the scatterer trajectory model under the steady motion premise of the target during sequential ISAR imaging.Then to deal with the complex trajectory observation situation,the Multiple Hypothesis Tracking(MHT)algorithm is introduced into the association process.And an improved MHT trajectory association method is proposed that estimates the trajectory motion parameters and the trajectory association state alternately.Compared with the traditional method,the proposed one eliminates the contradiction between the unknown trajectory model parameters and the requirement of known parameters for trajectory prediction.Furthermore,by matrix factorization,the 3-D distribution of target scatterers is reconstructed from the trajectory matrix.Finally,the effectiveness of the proposed method is verified by simulation experiments.3.Feature extraction from ISAR image sequence and attitude estimation for space targets The attitude information of the space target is implicit in the ISAR image sequence.However,because of the unclear mapping relationship between attitude parameters and ISAR image sequence,and the sparse,incomplete target structure in ISAR images,it is difficult to accurately extract the sequential image information and effectively apply it to attitude estimation.To solve this problem,this paper firstly analyzes the imaging geometry of the steady motion target,and accurately establishes the mapping relationship between the target attitude and the imaging results.In this way,we can generate the target projection image sequence under the given model and attitude parameters.Then,in the specific motion form of three-axis stabilization,the image energy accumulation value in sequential target projection area is adopted as the optimization variable.And without extracting the target structure features,the optimal attitude parameters are solved by maximizing the energy accumulation value.Further,in the general motion form of slow rotation around fixed axis,the Intersection over Union(Io U)of the target projection area and the extracted target area is adopted as the optimization variable.And the pix2 pix network is used to accurately extract the target area of the image sequence.By maximizing the Io U,the optimal target attitude parameters are estimated.Finally,by attitude estimation simulation experiments with the target in three-axis stabilization and fixed-axis slow rotation,the effectiveness of proposed methods is verified.4.Instantaneous state estimation for space targets with complex motions utilizing joint optical-and-radar observationUnder long-term observation,the state of space targets with complex motion is difficult to analyze by parametric model.To solve this problem,this paper firstly establishes a joint observation model of multiple optical and ISAR observation stations for space targets.And the instantaneous motion of the target at the observation instant is approximately modeled as a fixed-axis slow rotation.Specifically,the instantaneous state of the target is described by the component size,orientation and the target rotation vector.Then we analyze the multiview imaging geometry of joint multi-station observation at the same instant.And the spatial relationship between the optical image features and the ISAR image features is established.Further,comprehensively utilize the projection length of components in multi-view optical and ISAR images to reconstruct their 3-D size and instantaneous attitude.And the target rotation vector is estimated utilizing Doppler spectral width of components in the multi-view ISAR images.Finally,by simulation experiments,the effectiveness of the proposed method to solve the complex moving target state inversion problem is verified.