【作者】 吴迪；

【导师】 朱岱寅；

【作者基本信息】 南京航空航天大学 ， 通信与信息系统， 2011， 博士

【摘要】 作为一项先进的雷达信号处理方案，空时自适应处理（STAP）技术根据接收信号中杂波及噪声的统计特性自适应生成空、时二维滤波器，能够有效抑制地杂波，极大地提高了机载动目标指示（MTI）雷达以及合成孔径雷达地面动目标指示系统（SAR/GMTI）的慢动目标指示能力。然而，由于常规的STAP处理中需要采用与待检测单元杂噪分量独立同分布（I.I.D）的训练样本对杂噪协方差矩阵进行估计，而在实际的非均匀环境中，样本的I.I.D条件无法满足，从而导致了STAP性能的下降。针对这一点，本文研究围绕着各种非均匀STAP及SAR-STAP技术展开，主要内容及创新点如下：1研究了机载雷达地杂波模型，对杂波特性进行了分析，介绍了STAP全维最优处理器的结构，引入了STAP性能衡量的指标，对经典的降维3DT、降秩最小范数特征对消器（MNE）以及对角加载样本矩阵求逆（LSMI）算法进行了回顾。并在此基础上，通过对机载多通道雷达实测数据的分析与处理，对相关理论、算法进行了验证。2研究了非均匀STAP技术。重点对非均匀检测器（NHD）、空时自回归算法（STAR）以及结构化STAP算法进行了研究。首先将NHD引入降维处理中，采用基于广义内积（GIP）的反复剔除（RC）3DT算法对实测数据进行了处理，并获得良好效果。随后对STAR算法进行了研究，针对非平稳杂波过程造成的算法失效，提出了一种基于时变自回归（TVAR）模型的时变空时自回归（TV-STAR）算法，并通过仿真实验及实测数据的处理对其性能进行了验证。最后对结构化STAP算法进行了研究，针对原有算法中结构化矩阵不准确造成的性能损耗，提出了一种新的结构化STAP算法。新算法采用先验辅助（KA）方法构造杂噪协方差矩阵，通过加权调整使其逼近真实值，并由此产生最终的二维权矢量。实测数据处理结果表明，其性能优于原有的结构化STAP算法。3对自适应杂波抑制干涉（ACSI）SAR/GMTI技术进行了研究。ACSI技术仅需两个通道便能实现杂波抑制，是目前工程条件下最为实用的一种SAR-STAP技术。针对非均匀环境对ACSI算法造成的性能损耗，分别提出了条件约束ACSI算法及基于中值估计的中值对消ACSI算法，分别通过实测数据的处理对二者进行了验证。随后对CSI-SAR系统的运动目标参数估计方案进行了研究，给出了仿真及实测数据的处理结果。最后根据目前实际工程应用的需要，对三通道CSI-SAR的GMTI整体信号处理流程进行了设计，并给出了相应的实测数据处理结果。4研究了多通道图像域SAR-STAP技术。首先在Ender提出的后多普勒SAR-STAP（MSAR）技术的基础上，引入了图像域SAR-STAP技术。随后针对非均匀环境对其性能的影响，分别提出了两种非均匀图像域SAR-STAP算法，即先验辅助SAR-STAP算法和用于多通道SAR系统的改进FRACTA算法，并通过实测数据对两者的性能进行了验证。最后对SAR-STAP系统的动目标运动参数估计方案进行了研究，分析了极大似然估计（MLE）及自适应单脉冲测向法的性能，并对实测数据进行了处理。5作为全文中内容相对独立的部分，论文最后一部分对单脉冲成像技术进行了研究。提出了一种单脉冲成像算法，有效改善了机载/弹载雷达前视等SAR、多普勒波束锐化（DBS）技术成像盲区图像的清晰度。并从单脉冲和差比的概率密度函数出发，对单脉冲成像效果进行了分析，提出了目标图像位置失真、分辨率、图像信噪比三个图像质量衡量指标。最后，通过仿真实验及机载雷达实测数据的成像处理对算法性能进行了验证。更多还原

【Abstract】 Space-time adaptive processing (STAP) is a leading radar signal processing technique thatadaptively forms a two-dimensional filter with respect to the statistical characteristics of the clutterplus noise to effectively suppress the clutter component contained in the receiving data. It greatlyenhances the detection performance of airborne moving target indication (MTI) radar and syntheticaperture radar ground moving target indication (SAR/GMTI) system. However, for conventionalSTAP, a key step is to adaptively estimate the clutter plus noise covariance matrix, in which thetraining samples must be independent identically distributed (I.I.D) with the clutter plus noisecomponent in the cell under test. Unfortunately, in practical processing, the clutter environmentsalways present to be heterogeneous. That means the I.I.D assumption will never be valid, whichfinally leads to performance loss of STAP. Therefore, the major work in this dissertation focuses onthe research of STAP and SAR-STAP algorithms in nonhomogeneous environment and can besummarized as follow:The clutter model for airborne radar is studied and the clutter characteristic is analyzed.Fundamental theories such as optimal STAP weight vector as well as several criteria to evaluate theperformance are introduced. Classical reduced-dimension/rank algorithms such as3DT, minimumnorm eigencancerler (MNE), and loaded sample matrix inversion (LMSI) are investigated.Experimental results with respect to measured data collected by a three-channel airborne radar areemployed to verify the related theories and algorithms.The nonhomogeneous STAP algorithms are studied. Three nonhomogeneous algorithms, i.e., thenonhomogeneous detector (NHD), space-time autoregressive (STAR) filtering and the structuredSTAP are investigated, respectively. At first, the NHD is introduced into reduced-dimension STAPprocessing by employing a generalized inner product (GIP) based reiteratively censored (RC)3DTalgorithm to process the measured data, which is proved to be of great improvement. Then, aiming atthe failure of STAR in nonstationary clutter in slow-time, a new time-varying (TV) STAR algorithmthat based on the time-varying autoregressive (TVAR) model is proposed. The virtue of TV-STARrelative to STAR is verified by simulation as well as experimental results. At last, according to theperformance loss caused by the inaccuracy of the so called structured covariance matrix, a newstructured STAP algorithm is proposed. In the new proposed algorithm, the clutter plus noisecovariance matrix is formulated by knowledge-aided (KA) method and undergoes several adjustmentsvia tapering matrices. After that, the adjusted matrix is finally employed to calculate the space-time weight vector. The desirable detection performance is also verified by experimental results.The adaptive clutter suppression interferometer (ACSI) SAR technology is investigated. The socalled ACSI technique that introduces adaptive processing in CSI-SAR system cancels the cluttercomponent effectively by two receiving channels, and is considered to be the most valuableimplementation of SAR-STAP in practical processing. According to the performance loss caused byheterogeneous environment, two nonhomogeneous ACSI algorithms, i.e., the constraint ACSI andmedian canceller (MC) ACSI algorithm based on median estimation method are proposed and verifiedby the experimental results. Then, the parameter estimation methods for the detected moving target inCSI-SAR system are studied, which is also supported by simulation as well as experimental results.Finally, according to the practical engineering application, an entire CSI-SAR/GMTI signalprocessing scheme is designed for the three-channel SAR system and some related experimentalresults are provided.The image domain multi-channel SAR-STAP technology is studied. Based on the post-DopplerSAR-STAP method (MSAR) proposed by Ender, the image domain SAR-STAP technique isintroduced to detect moving target after SAR imaging. Then two nonhomogeneous SAR-STAPalgorithms, i.e., knowledge-aided SAR-STAP and modified FRACTA for multi-channel SAR, areproposed to operate in heterogeneous clutter environment and also verified by experimental resultswith respect to the three-channel SAR system. At last, the parameter estimation methods formulti-channel SAR system are investigated, the maximum likelihood estimation (MLE) as well as theadaptive monopulse method is introduced. The performance of each method is analyzed bysimulations and verified by experimental results.As a relatively independent part, the monopulse imaging technology is investigated at the end ofthis dissertation. A monopulse imaging algorithm is proposed to improve the definition of radar imageconcerning the airborne/missle-borne radar forward-looking area, which can be hardly achieved bySAR or Doppler beam sharpening (DBS) technique. Furthermore, based on the probability densityfunction of the monopulse ratio, three criteria, i.e., the image position distortion, resolution andsignal-to-noise ratio of the target to be imaged are proposed to evaluate the quality of monopulseimage. Factors determining the proposed criteria are analyzed afterwards, followed by the simulationand experimental results to verify the performance of the proposed algorithm.更多还原