【作者】 邓彬；

【导师】 黎湘；

【作者基本信息】 国防科学技术大学 ， 信息与通信工程， 2011， 博士

【摘要】 目标微动往往蕴含着对合成孔径雷达(SAR)图像解译极为有利的特征和信息,同时也会造成目标SAR成像模糊等不利影响,且传统的SAR地面运动目标指示(SAR/GMTI)技术无法解决。本文以SAR复杂运动目标探测为背景,系统研究目标微动对SAR和SAR/GMTI的影响以及微动目标检测、参数估计和成像问题。第二章研究SAR微动目标与背景建模。建立了目标转动、振动、正弦运动和摆动四种微动模型,以及理想点、展布式、局域式和滑动型四种散射中心模型,提出了旋翼两种散射中心结构,建立了抛物面天线滑动型散射中心模型。给出了基于2D FFT的SAR背景回波快速生成方法适用的充要条件,提出了斜视匀直航迹、匀加速航迹和误差航迹SAR回波快速生成方法,解决了背景回波生成效率低的难题。提出了基于Legendre多项式展开的SAR成像新方法,具有相位近似误差小、适应斜视角大等优点。第三章研究目标微动对SAR和SAR/GMTI的影响。发现了反映目标各阶运动参数与相位误差关系的锯齿现象、小幅快速微动的距离徙动敏感性现象和微动目标PFA图像张角效应。提出了锯齿原理和广义成对回波原理,以此为理论基础分析了不同微动类型、参数、目标个数、散射类型对不同成像算法和SAR极限方位分辨率的影响,揭示了微动目标SAR图像的八类典型特征。指出目标微动将严重恶化单通道SAR杂波抑制和目标检测性能,但对多通道杂波抑制方法影响较小。第四章研究杂波抑制和微动目标检测问题。提出了基于回波广义似然比(GLRT)的SAR一般微动目标检测方法,从理论上推导了检测器性能。对于大幅微动,提出了基于分段检测积累和断续正弦曲线特征的微动目标检测方法,通过相干-非相干混合积累获得了高信噪比增益。对于小幅微动,提出了基于SAR图像鬼影特征的微动目标检测方法,将脉冲重复周期变换应用于鬼影检测,具有不受多倍周期误差影响等优势。此外还提出了基于偏置相位中心天线(DPCA)的双通道杂波抑制和微动目标检测方法,实现了杂波频带内的微动目标检测。上述算法同时具有一定的参数估计能力。第五章研究微动目标重新聚焦成像,包括微动目标距离徙动校正(RCMC)、匀速平动分量补偿、动静目标信号分离、微动目标多普勒解缠、中心频率估计和成像等问题。提出了SAR微动目标需要RCMC的准则,以及基于降带宽思想、相位补偿和多普勒域级联Keystone变换的RCMC方法。对于微动目标所在距离单元,基于正弦调频(SFM)信号特点提出了匀速平动分量补偿的Wigner-Hough变换方法。采用Chirplet分解方法实现了多普勒混叠下的微动-残余静止目标回波的分离。对于方位聚焦,提出了基于序列插值的多普勒解缠方法以恢复微动正弦相位和时频曲线,并采用能量均衡法估计SFM信号中心频率,在此基础上提出了消除微动的目标自聚焦成像和利用微动的目标逆Radon变换伪窄带成像方法,分析得出各向异性散射导致逆Radon变换成像结果只能反映散射系数平均值的结论。第六章研究了SAR微动目标检测-成像联合实现技术。首先提出了SAR目标检测-成像联合实现理论框架,包括正问题建模、初值估计/先验建模、逆问题求解三大部分。对任意运动任意散射目标SAR回波用第一类Fredholm方程建立正问题观测模型,对于微动目标又具体化为两种运动-散射混合模型——微动-点散射中心模型和微动-滑动型散射中心模型。接着详细分析了微动目标RCS、距离像、时频分布和二维像特性,发现了旋转抛物面天线SAR图像“蝴蝶结”特征,提出了参数初值估计方法,为逆问题优化求解算法快速全局收敛提供了保证。最后提出了从微动目标层析和参数估计角度对逆问题进行求解的思路,重点围绕参数估计提出了微动目标观测模型优化求解的最大似然估计方法,实现了运动-散射参数联合估计。该框架最大程度地利用了目标模型先验和目标特性,通过引入目标模型先验自动隐含了目标检测与成像,并具有集成运动补偿和对动静目标联合清晰成像的潜力。总之,本文在概念、模型、现象、原理、特征、思想和方法等方面对传统SAR/GMTI技术进行了推广,揭示了微动目标SAR图像特征及其形成机理,提出了一系列微动目标检测、成像方法和检测-成像联合实现理论框架,实现了微动目标成像和在SAR图像上的标注。取得的成果对于高分辨对地观测、精细化探测、SAR海量图像快速解译和干扰对抗具有一定的参考价值。更多还原

【Abstract】 Target micro-motion conveys features and information which are favorable for understanding synthetic aperture radar (SAR) images. However, micro-motion will result in defocusing and other unfavorable effects on target imagery, which can’t be overcome by conventional SAR/ground moving target indication (SAR/GMTI). Therefore, this dissertation aims at micro-motion targets in SAR and detailedly investigates their detection, parameter estimation, and focusing approaches.Chapter 2 focuses on models of SAR micro-motion targets and backgrounds. We develop four types of micro-motion models including rotation, vibration, sinusoidal motion and rocking, as well as four types of scattering models including ideal point, distributed, localized and migratory scattering. We propose for blades two scattering center structures, and build a migratory scattering center model for parabolic-reflector antennas. Then background models and SAR echo simulation are studied. We derive a sufficient and necessary condition for applying 2D FFT methods to generate raw echos, and propose three kinds of algorithms for squint looking SAR with a constant velocity, constant acceleration and trajectory deviation, respectively, which enhance the simulation efficiency particularly for large backgrounds. Based on the resultant echoes, a novel approach to range Doppler SAR echo processing based on Legendre orthogonal polynomials is presented with smaller phase approximation errors and improved squint-looking applicability.Chapter 3 examines micro-motion effects on SAR and SAR/GMTI. We discover the sawtooth phenomenon between motion and phase orders, the range cell migration (RCM) sensitivity phenomenon incurred by minor-amplitude and fast-frequency micro-motion, and the angular-extent effect of micro-motion targets on images by the polar format algorithm. The sawtooth principle and the generalized paired echo principle are proposed. Then taking this as the rationale, we analyze in detail the effects of micro-motion types, parameters, target numbers and scattering on SAR algorithms and on the resolution limts. Eight typical kinds of features of SAR imagery are revealed, and we also conclude that micro-motion will considerably degrade the detecting performance for single-channel SAR, while has little influence on that of multi-channel SAR.Chapter 4 investigates clutter suppression and micro-motion target detection. A generalized likelihood ratio test (GLRT) detector are developed using SAR returns as opposite to images, the detecting performance are derived theoretically. Then for major-amplitude micro-motion, a special detection method is proposed which uses post-detection integration and discontinuous sine curve characteristics to obtain a high signal-to-noise ratio gain. Also for minor-amplitude micro-motion, an alternative based on target ghost images are proposed which uses the pulse repetition interval (PRI) transform to detect the ghost points and dispenses from the multiple period error. In light of limited single-channel performance, we also suggest another approach, based on dual-channel displaced phase center antenna (DPCA), for rejecting clutters and detecting micro-motion targets within the clutter spectra. These detection methods also have the capability of estimating micro-motion parameters.Chapter 5 concentrates on the refocusing of micro-motion targets, including translation compensation, RCM correction (RCMC), micro-motion signal separation/extraction, micro-Doppler unwrapping, Doppler centroid estimation and imaging. A norm on when RCM must be corrected is at first proposed. Then three RCMC algorithms are put forward based on bandwidth reduction, phase compensation and the Doppler successive Keystone transform, respectively. For range cells containing micro-motion targets, a Wigner-Hough-transform based method is proposed to compensate the translation, which uses characteristics of sinusoidal frequency-modulated (SFM) signals. In what follows, we use adaptive Chirplet decomposition to separate micro-motion returns from stationary ones, and consider the effect of Doppler aliasing on decomposition. For target azimuthal focusing, a sequence-interpolation based Doppler unwrapping approach is proposed for recovering the sinusoidal phase and time-frequency curve. Afterward the energy balancing method is used to estimate and compensate the Doppler centroid of SFM signals, based on which a micro-motion-eliminated autofocusing method and a micro-motion-utilized imageing method via inverse Radon are suggested respectively. We also conclude that, for anisotropic targets, their image by the inverse Radon transform only reflects the averaged scattering intense.Chapter 6 discusses the joint detection and imaging technique for SAR micro-motion targets. We at first propose a theoretical framework for joint detection and imaging of SAR target, including (1) forward problem modeling, (2) initial value estimation and prior information modeling, and (3) inverse problem solving. For (1), we prove that SAR target returns, with arbitrary scattering and/or motion, can be modeled by a Fredholm integral equation of the first kind. For our micro-motion targets, the equation is embodied by a hybrid motion-scattering model, i.e. the micro-motion/point-scattering model and the micro-motion/migratory-scattering model. For (2), we analyze radar cross section, range profiles, time-frequency distribution and 2D image characteristics of micro-targets, and reveal the bowknot-shaped feature of rotating parabolic antennas. Based on this, the initial value estimation method is given which can guarantee fast global convergence when solving the inverse problems. For (3), we propose that the inverse problem can be solved from the viewpoint of either the micro-motion target tomography or parameter estimation, and particular emphasis is placed on the latter one. A maximal likelihood estimation method for solving micro-motion target models is proposed to realize joint estimation of both motion and scattering parameters. This framework exploits prior information on target models and their characteristics to the utmost extent, and implicitly incorporates detection and imaging by introducing target model prior information, while also offers the potiential of incorporating platform motion compensation and jointly imaging both moving and stationary targets.To summarize, we have in this dissertation uncovered micro-motion target image characteristics and their formation mechanism, proposed a series of detection and imaging algorithms as well as a theoretical framework for joint detection and imaging. This dissertation has realized micro-motion target imaging and indication on SAR images. These research results bear certain significance for high-resolution and refined earth observation, precison bombing, fast SAR image interpretation and passive SAR jamming as well as its countermeasure.更多还原