High-accuracy SAR Target Reconstruction Based On Partial Scattering Model

Synthetic Aperture Radar(SAR)has the ability of acquiring the target information under the condition of all weather and all day.Advances in radar hardware and electronic system have enabled the modern SAR system with the multitasking capability such as high-resolution image formation,high-precision target feature description and autofocus,etc.Meanwhile,The applications in both the military and civil field have pressing needs and great expectations of achieving the target higher resolution and more detailed description.In order to improve the ability of SAR reconstruction and help radar automatic target recognition,SAR target high-precision reconstruction algorithms based on the partial scattering model of the target are studied in this dissertation.The partial scattering characteristics of the target are the spatial sparsity of the scattering centers,the anisotropic scattering and object-level parameterized scattering.The main work in this dissertation can be shown as follows.1.Electromagnetic scattering mechanism-driven SAR object-level reconstruction methodIn Chapter 3,the relationship between the object-level parameterized scattering model and SAR observation has heen investigated from the point of view of the electromagnetic scattering mechanism.Firstly,in sparse representation framework,the object-level SAR observations are modeled based on attributed scattering center(ASC)model and canonical shape feature(CSF)model.The following work has been done for SAR object-level reconstruction by the strategy design of dimension reduction and the formula derivation.To mitigate the problem of the weak scattering center reconstruction failure due to the high energy difference between the weak and strong scattering centers,the ASC based PRS-ROMP(Peak Region Segmentation and Regularized Othogonal Matching Pursuit)reconstruction algorithm is proposed.The weak and strong scattering centers are reconstructed by stages.The strong scattering center attributes are reconstructed by PRS and least square(LS).Then,the weak scattering center attributes are reconstructed by ROMP within the residual signal.Experimental results of the synthetic scene data and electromagnetic calculations data of the simple tank validate the proposed algorithm.Aiming at the problem that PRS in image field suffers from unsuitability to reconstruct the distributed scattering center with higher precision.We propose a classifying strategy of the target attributes space for the object-level reconstruction in signal domain.Combined with data extrapolating,a stochastic gradient minimum variance pursuit(SGMVP)based object-level superresolution reconstruction algorithm is proposed,which not only can achieve improved superresolution image but also provide accurate physically-relevant attributed features of the scatterers simultaneously.Experimental results confirm the effectiveness of the proposed algorithm.A CSF based object-level SAR sparse reconstruction algorithm is proposed.We develop a space-decomposition strategy to avoid the curse of dimensionality of over-complete dictionary,meanwile,make full use of the advantages of the regularization technique in prior information fusion,design regular term to mitigate the interferences between different spatial locations,derive the solving equations by means of quasi-newton optimization process and prove the consistency of real-valued gradient and the compact form of SAR complex-valued gradient.2.Adaptive high resolution wide-angle SAR reconstruction methodIn Chapter 4,considering the anisotropic scattering characteristics of the target under the wide angle observation,we peopose a full-aperture adaptive wide-angle SAR reconstruction algorithm based on MAP criterion.First,the full-aperture wide-angle SAR reconstruction model is built according to the scattering dependency of both the spatial location and the viewing angle.We model the directional selectivity and high azimuth correlation of scattering energy by virtual of a special Boltzmann machine(BM)prior.Then,the support of sparse representation and algorithm parameters including BM parameters,noise variance and the variance of each sparse representation element are jointly estimated by a block-coordinate descent process.The proposed algorithm can obtian wide angle composite SAR image with high resolution and the anisotropic scattering characteristic curve of the target with high precision simultaneously.3.New type rotating arm arc spotlight SAR reconstruction methodIn Chapter 5,ground-based new type rotating arm arc spotlight SAR(RAAS-SAR)reconstruction autofocus is mainly studied.First,the principle of RAAS-SAR system is introduced,and the high resolution in both range and cross range are analyzed.Then,the RAAS-SAR observation model is built.Finally,based on the geometry and reconstructing principle of RAAS-SAR,the adjusted version of the Range-Doppler reconstruction formulas are derived.The simulation results validate feasibility of the RAAS-SAR system and the proposed algorithm.For the complex white Gaussian noise and the phase noise coexist circumstances,combined the spatial sparsity,a quadratic compressed sensing(QCS)based SAR reconstruction autofocus algorithm is proposed.We recast the RAAS-SAR reconstruction problem via the phase-corrupted data as for a special case of QCS.Although the quadratic measurement model has potential to mitigate the effects of the phase noises,it also leads to a nonconvex and quartic optimization problem.A variant of 2-opt search technique and damped Gauss-Newton are used to update the support and the scattering coefficient vector iteratively.In addition,a modified fast reconstruction autofocus algorithm is proposed by means of the block decomposition.A adaptive RAAS-SAR sparse reconstruction autofocus algorithm is proposed based on evidence maximization framework.A specific Gaussian scale mixture prior in the Power Exponential Scale Mixture(PESM)family is used to impose sparsity constraint,meanwhile,the Von Mises(VM)prior in directional statistics is exploited to model the circular statistic characteristics of the pulse-index phase noise.The update equations of the sparse representation,phase noise matrix and hierarchical algorithm parameters are derived.The proposed algorithm can obtain the satisfied estimation of pulse-index phase noise and the well autofocus results.