【作者】 张磊；

【导师】 保铮；

【作者基本信息】 西安电子科技大学 ， 信号与信息处理， 2012， 博士

【摘要】 高分辨合成孔径雷达和逆合成孔径雷达(SAR/ISAR)成像技术具有全天候、全天时和远距离成像的特点，有效提高了雷达的信息获取能力，具有重要的军用和民用应用价值。SAR/ISAR成像中，分辨率的提高对精细表征观测目标至关重要。距离分辨率通过发射带宽信号获得，方位分辨率则取决于合成孔径大小。在SAR/ISAR应用中，提高二维分辨率不仅受雷达体制的制约，也对合成孔径的阵列误差更加敏感，需要更为精确稳健的运动补偿。通过合理的频带和时间资源分配，结合相控阵技术的多功能ISAR具备广域、多目标成像的能力。但在多目标探测中，对单一目标的频带和孔径观测将是稀疏有限的，这在信号处理中需要加以克服。结合现代无人机等小型化平台的高精度SAR具有很强的灵活性和机动性，是现代SAR发展的一个重要方向，但低空小型平台对大气扰动更为敏感，且难以配备高精度惯性导航系统进行运动补偿，在成像处理中亟需稳健高效的自适应运动补偿技术。本论文旨在利用信号处理方法提高SAR/ISAR成像的分辨率、探测区域、灵活性和稳健性，研究内容主要针对目标超分辨成像、稀疏频带和稀疏孔径高分辨成像、精确稳健的自适应运动补偿和多通道SAR宽域高分辨成像四个关键点。论文围绕国家“973”计划课题“稀疏微波成像的理论、体制和方法研究”、国家自然科学基金重大项目“多维度微波成像基础理论与关键技术”以及“863”课题“空间目标雷达宽带特性测量与成像研究”等项目的研究任务，对高分辨SAR/ISAR成像和误差补偿方法进行了研究。全文内容主要针对目标超分辨成像、稀疏频带和稀疏孔径成像，稳健精确的SAR运动补偿自聚焦和多通道SAR成像四个方面，概括为以下四个部分：第一部分研究基于稀疏重建的目标超分辨成像。建立了基于稀疏重建理论的超分辨成像的一般模型，分析了影响稀疏重建超分辨的若干重要因素及其确定方法。针对低信噪比情况，通过构造加权因子以区分目标信号支撑区和背景，提出了基于改进压缩感知的超分辨重建方法。从贝叶斯压缩感知出发，建立了范数1正则化超分辨成像的一般模型，推导了正则化优化函数中范数权系数的含义及其最大似然估计表达。通过引入同分布和非同分布统计模型，分别提出了贝叶斯超分辨成像和改进贝叶斯超分辨成像算法。建立了分步迭代估计统计参数和超分辨成像重建的处理流程，建立了结合快速傅立叶变换的改进柯西-牛顿求解算法。在此基础上，结合稀疏重建的超分辨算法提出了短孔径ISAR成像、机动目标ISAR成像等多种实用方法，有效提高了ISAR目标成像质量。第二部分研究稀疏频率和稀疏孔径的高分辨成像。建立了稀疏步进调频信号的高分辨距离像重建优化求解算法。针对稀疏步进调频ISAR运动补偿，结合包络偏移估计、自聚焦以及多频多普勒速度估计等方法提出了稳健精确的统计参数和运动参数估计流程。从贝叶斯统计理论出发，建立了稀疏孔径ISAR成像算法。针对稀疏孔径间存在非连续运动误差，建立了联合高分辨成像和初相校正的优化求解方法，还提出了结合全极点信号模型的稀疏孔径相干化处理方法。所提出的方法改善了ISAR成像雷达的频带和时间资源利用率。第三部分研究基于扩展相位梯度自聚焦的SAR自适应运动补偿方法。在传统的相位梯度自聚焦算法(PGA)的基础上，提出了局部最大似然-加权相位梯度自聚焦（LML-WPGA）算法，实现对距离空变运动误差精确估计。针对条带式SAR运动补偿，提出了基于WPGA和LML-WPGA的自适应运动补偿方法。该方法分步校正包络偏移误差、非空变相位误差以及空变相位误差，并结合重叠子孔径和低通滤波技术实现条带模式下的高精度全孔径自适应运动补偿，研究中算法还被推广到了大斜视SAR成像处理中。将LML-WPGA的思想推广到两种现有的自聚焦算法(PWE-PGA和WPCA)，建立了LML-PWE-PGA和LML-WPCA算法，极大改善了算法的运算效率和精度。第四部分研究多通道SAR宽域高分辨成像和通道均衡。在利用多接收通道对空域解多普勒模糊方法系统分析的基础上，提出了基于稳健波束形成的解多普勒模糊成像方法。相比传统的自适应解模糊，基于稳健波束形成的解模糊方法在抑制多普勒模糊分量的同时，实现自适应搜索目标信号的真实导向矢量，有效提高了解模糊算法对多通道SAR信号幅相误差的容忍性。针对存在较大通道误差的情况，提出了距离和方位分维误差校正的自适应通道均衡方法，建立了子空间信号处理的幅相误差估计方法，该方法可有效利用多个甚至所有多普勒单元信号联合估计方位维非空变和慢空变的幅相误差，并在子空间投影中利用天线方向图加权等技术，有效改善了通道误差估计的精度。更多还原

【Abstract】 High resolution synthetic aperture radar and inverse synthetic aperture radar (SAR/ISAR)imaging technique has the ability of well-weather, day/night and long range applications,which dramatically enhance the capability of information acquisition of modern radar.Therefore, SAR/ISAR technique plays an essential role in many military and civilian fields.In SAR/ISAR imaging, high resolution is very important to represent the detailedcharacteristic of the target. Range resolution relies on the bandwidth of the transmitted signal,and azimuth resolution depends on the synthetic aperture size. Two-dimension resolution islimited by not only the radar system constraints both also the manifold accuracy of thesynthetic aperture. Generally, with the increase of azimuth resolution, focusing performancewould be very sensitive to the systemetic errors, such as motion errors, demanding precisecompensation schemes. By arranging the radar frequency and time resource suitably andcombining with phased array technique, modern multi-function ISAR accomplishes multiplytasks simutantaneously, such as wide-swath surveillance, multi-target tracking and imaging.However, in this case, the frequency band and aperture for a single target is limited and sparse,which should be accounted in the imaging processing. Owing to its fleasibility andmaneuverability, compact SAR mounted on small platforms, such unmanned aerial vehicleand missile, is very important for modern battlefield survelliance. However, because of itssmall size and light-load capability, it is sensitive to the atmospheric turbulence, andfurthermore, the high-precision inertia navigation system is usually unavailable due to loadcapability constraint of the platforms. Therefore, prcesise and robust motion compensationbased on raw data is desiderated.This dissertation studies new techniques to improve the resolution, operational swath,feasibility and robustness of SAR/ISAR from four key aspects, i.e. resolution enhancementwith sparse representation, high resolution imaging from sparse frequency bands andapertures, precise and robust motion compensation based on raw data, andhigh-resolution-wide-swath imaging with azimuth multi-channel SAR. The relevant work issupported by by the National Basic Research Program of China （973Program, No.2010CB731903）, National Science Foundation of China (No.60802081and No.60890072)and the National High Technology Research and Development Program of China (No.2008AA8080402).The main content of this dissertation is summarized as follows.The first part focuses on the super-resolution imaging based on sparse representation. A general compressive sensing (CS)-based imaging scheme is built. Some related factorsaffecting the performance of the algorithm are analyzed in detail, based on which we alsodevelop the approach to estimating the related parameters. Accounting for the strong noise,the signal and noise supports are distigushed via introducing optimal weigths, and theimproved CS super-resolution imaging method is proposed. Developed from Bayesiancompressive sensing (BCS), we build the norm1-regularition-based super-resolution imagingscheme, and the weighting parameter for norm1term and its maximum likelihood estimationare derivated mathematically. Non-identical statistics model is extended to the BCS-basedoptimization, and two super-resolution algorithms, Bayesian super-resolution and improvedBayesian super-resolution (BSR and IBSR), are developed. A stage-by-stage procedure isdeveloped to jointly estimate statistics parameters and reconstruct the super-resolution image.Combining with fast Fourier transform, we also propose a modified Quasi-Newton solver toBSR and IBSR optimizations. Some applicable approaches for short-aperture ISAR imagingand maneuvering target imaging are also developed. The validity of the proposed methods isproved by several sets of real-measured data.The second part studies high resolution radar imaging by exploiting sparse frequencybands and apertures. High resolution range profile synthesis by sparse representation and therevelant parameter selection are developed. For ISAR imaging with sparse stepped-frequencywaveforms, precision motion compensation procedure by combining optimal range alignment,autofocusing and parameter estimation with multi-frequency diversity is developed. Based onBayesian compressive sensing, high resolution imaging with exploiting sparse apertures isdeveloped. In the algorithm, the discontinuous phase error function is overcome by sparseaperture coherence processing, which can be jointly implemented in the imaging optimization.In terms of high precision and efficiency, a pre-processing for phase error correction ispresented. Extending from the all-pole model, we develop a novel coherent processing tocorrect the linear and constant phase difference between sub-apertures. Real data sets areutilized to confirm the validation of the proposed methods.The third part presents the SAR motion compensation (MOCO) based on the extendedphase gradient autofocus (EPGA). In this part, we develop the local maximumlikelihood-weighted phase grandient autofocus (LML-WPGA) algorithm, which is capable ofprecsion estimation of range-dependent phase errors. In terms of precise MOCO for thestrip-map SAR, a procedure implemented by weighted phase gradient autofocus andLML-WPGA is proposed, which corrects nonsystematic range migraition, nonspatial-variantand spatial-variant phase errors sequently. Subaperture overlapping and adaptive filtering areutilized to construct full-aperture motion error function from raw data. The MOCO scheme is extended to highly squinted and spotlight SAR imaging. And LML is also introduced into twoautofocuse PWE-PGA and WPCA. Therefore, LML-PWE-PGA and LML-WPCA aredeveloped, which have the propertities of high precision and efficiency. The proposedalgorithms are validated by using a number of real SAR sets.The last part focuses on high-resolution-wide-swath (HRWS) imaging withmulti-channel SAR and adaptive channel calibration. Based on detailed analysis on theDoppler ambiguity resolving with spatial filtering beamforming, the robust beamformingtechniques are introduced into multi-channel SAR imaging. Comparing with conventionalDoppler resolving method, robust beamforming-based method can not only suppressambiguity components adaptively, but also precise reconstruct signal component by arrayvector estimation. By the robust beamforming technique, the performance of multi-channelSAR imaging is enhanced effectively. A two-step channel calibration is proposed based on thefact that the amplitude and phase errors between channels are usually uncoupling in range andazimuth directions. Subspace-based calibration is developed to correct the channel mismatch,whose performance and robustness are ensured by using multiply Doppler bins and weightingsubspace projection. Two sets of real measured data are utilized to confirm the effectivenessof the proposed methods.更多还原