| Spaceborne synthetic aperture radar(SAR)has been widely used in military reconnaissance,civil construction and scientific research because of its day and night,all-weather,and global observation capabilities.In order to achieve refined resolution,high-precision geolocation,and all-directional description of targets,a novel spaceborne all-directional SAR system has been proposed in recent years.The SAR satellites fly around the same ground area with four different orbits,and by controlling the radar beam to achieve all-directional observation.During each pass,the radar beam has a squint angle in the range of-45 to 45 degrees,and the synthetic aperture time can reach an order of hundred seconds.In this dissertation,the three-dimensional(3D)reconstruction technologies for the spaceborne all-directional SAR system have been studied from two aspects including non-coherent processing and coherent processing.Most of targets have small coherent scattering angles.In order to obtain their SAR images with high signal to noise ratio(SNR),the full-aperture data is divided into several subapertures,and then the data of each subaperture is focused to form a raw image.Non-coherent processing uses the target's shifts between multi-angle SAR images to perform 3D reconstruction.For the target with large coherent scattering angle,the radar's trajectory forms a long synthetic aperture in 3D space during long-term observation.This makes the spaceborne all-directional SAR system has a 3D high resolving power.Coherent processing is to perform 3D reconstruct of the target with large coherent scattering angle by performing matched filtering on its echoes among the effective scattering angles.The main work of this dissertation is summarized as follows:(1)Basic concepts and principles of 3D reconstruction for spaceborne all-directional SAR In Chapter 2,the geometry of spaceborne all-directional SAR system and the parameters that describe its geometric relationship are introduced.For the multi-angle SAR image sequence,the relationship between the target's shifts and its height is given through theoretical derivation.For the target with large coherent scattering angle,the 3D resolution of spaceborne all-directional SAR system is given by the theoretical derivation.(2)Non-coherent 3D reconstruction for spaceborne all-directional SAR In Chapter 3 and Chapter 4,for the multi-angle SAR image sequence acquired from the spaceborne all-directional SAR system,the non-coherent processing method is studied.? Through theoretical derivation,analysis of simulated and real data,the characteristics of multi-angle SAR image sequence are analyzed.It is discovered that with the increase of azimuth angle interval of image pair,its correlation decreases.Based on this,the multi-angle SAR images are divided into correlated image pairs,which have small azimuth angle intervals,and non-correlated image pairs,which have large azimuth angle intervals.For the correlated image pairs,their shifts can be directly and accurately estimated;for the non-correlated image pairs,it is difficult to directly and accurately estimate their shifts.? For the multi-angle image sequence acquired by the spaceborne all-directional SAR system,the multi-angle SAR image sequence stereo radargrammetry method for 3D reconstruction is first proposed.The method comprises two steps: the first step is to perform coregistration processing on the multi-angle SAR image sequence.For the correlated image pairs,a terrain elevation adaptive method is proposed.In this method,the shifts are approximated by a two-dimensional(2D)three-variable polynomial of the first degree including the variable associated with the terrain elevation,and the first-order coefficients of the polynomial are calculated by the imaging parameters and orbital data.The precision and robustness of coregistration is improved.The precision and robustness of the proposed method are validated by real data.For the non-correlated image pairs,in order to solve the problem that their correlation is low and it is difficult to directly and accurately coregister,a least squares based joint coregistration method is proposed.The method utilizes the shifts of the correlated image pairs to estimate the shifts of the non-correlated image pairs.The simulated experimental results show that the accuracy of the shifts of the non-correlated image pairs obtained by this method and the correlated image pairs are of the same order.The second step is to construct and solve the range-Doppler equation set.The range-Doppler equation set that constructed based on the coregistered multi-angle SAR images is nonlinear,and it is difficult to obtain the analytical solution.A Newton iteration based method is proposed to this equation set.The simulated experimental results verify the effectiveness of the method.? In the aforementioned method,the coregistration of multi-angle SAR image sequence and resolution of range-Doppler equation set are cascaded.If the coregistration step results in an information loss,the next step cannot recover the lost information.To solve this problem,a joint coregistration and height estimation method is proposed.In this method,a system of linear equations is established based on the relationship between the shifts of correlated image pairs and non-correlated image pairs and the relationship between the shifts and the target's height.The least squares method is used to solve the system of linear equations,and the shifts and target's height are obtained simultaneously.The simulated experimental results show that the method avoids the transmission error introduced by the cascade processing and improves the accuracy and robustness of 3D reconstruction.? The joint correlation method is proposed.In this method,the joint correlation coefficient is used as a measure function.The joint correlation coefficients with the reset target's heights are calculated.The height corresponding to the maximum value of the joint correlation coefficient is treated as the estimated height.The experimental results based on real data show that the method makes full use of joint correlation information of the multi-angle SAR image sequence and improves the accuracy of 3D reconstruction.An efficient inverse geolocation method for high squint spaceborne SAR is proposed.The method is used to project the multi-angle SAR images onto the 3D space in order to correct their geometric deformation and improve their joint correlation.The method utilizes the characteristic that the differences between the shifts of the adjacent targets on the image and the differences between their three-dimensional positions satisfy the linear relationship.? The low-rank and sparse matrix decomposition method is proposed.First,each image is reshaped into a vector,and all of the vectors are stacked up,therefore a composite matrix is formed.Second,the shifts between the subspaces of the composite matrix are estimated by coregistration processing,and then the shifted subspaces are corrected.Finally,the corrected composite matrix is decomposed into a low-rank matrix and a sparse matrix in order to extract the targets with same height,and the height of the extracted targets is estimated from the shifts between the subspaces.The sparse matrix is treated as a new composite matrix and then performs the second and third steps iteratively until all of the targets are extracted.The method utilizes the fact that the shifts of the targets with same height on multi-angle SAR images are same.Therefore,it can simultaneously extract multiple targets with the same height and estimate their height with high-precision.The experimental results based on real data verify the effectiveness of the method.(3)Coherent 3D reconstruction for spaceborne all-directional SAR In Chapter 5,the coherent processing method is studied for the target with large coherent scattering angle.The spaceborne all-directional SAR system has 3D high resolving power for the target with large coherent scattering angle.Based on this characteristic,an adaptive 3D reconstruction method is proposed.First,the weighted echo inversion algorithm is proposed to extract the target with a large coherent scattering angle and its effective scattering angles.Second,a nonlinear equation set is constructed based on the signal model.Finally,the stage-wise subspace pursuit algorithm is proposed to solve the problem.The experimental results based on real data show that the method can automatically extract the target with large coherent scattering angle and perform high-precision 3D construction. |
PDF Full Text Request