【作者】 韦顺军；

【导师】 张晓玲；

【作者基本信息】 电子科技大学 ， 信号与信息处理， 2013， 博士

【摘要】 线阵三维SAR作为一种新型的三维雷达成像技术,在高精度测绘与资源调查、自然灾害监测与侦察预警等民用和军用领域都有广泛及重要的应用价值。受限于传统Nyquist采样定理和经典信号处理理论,目前线阵三维SAR存在成像分辨率过低、硬件系统实现困难、信号采样率过高、回波数据量大、数据的传输、存储以及处理困难等瓶颈难题。压缩传感稀疏重构作为一种近几年新提出的信号处理理论,突破了传统Nyquist采样定理约束,可利用远低于Nyquist采样率精确重构原始稀疏信号,在降低线阵三维SAR系统采样率和提高成像质量等方面有着巨大的应用潜力。本论文以压缩传感稀疏信号处理理论为核心,以高分辨率线阵三维SAR稀疏成像机理与方法作为研究内容,重点围绕线阵三维SAR稀疏成像技术中回波信号线性表征、稀疏重构成像方法、相位误差估计与补偿和线阵阵元分布优化等关键问题展开深入研究。论文的主要工作和创新总结如下：1.研究线阵三维SAR稀疏成像基本原理。从线阵三维SAR回波信号与成像空间映射关系入手,建立了线阵三维SAR距离向、阵列平面维和全场景三维空间回波信号的线性测量模型,将线阵三维SAR成像处理转化成散射系数线性方程求解问题,为线阵三维SAR成像新方法研究提供了理论基础；构建了基于压缩传感稀疏信号处理理论的三维线阵SAR稀疏成像处理的总体模型,分析了线阵三维SAR成像空间中目标散射系数稀疏表示、回波信号稀疏采样方式、复数域稀疏重构方法以及稀疏成像分辨率特性；针对线阵三维SAR大规模数据处理,提出了阵列平面与距离向分维处理的线阵三维SAR稀疏成像方法；通过理论推导,比较了传统匹配滤波方法、最小二乘方法和压缩传感稀疏重构方法的成像性能。2.研究了线阵三维SAR稀疏成像算法。将OMP稀疏重构算法应用于线阵三维SAR复数域稀疏成像,分析OMP算法稀疏成像性能；针对线阵三维SAR成像中目标稀疏度未知情况,提出一种硬阈值OMP稀疏成像算法,利用目标散射系数变化率作为算法迭代终止条件,未知目标稀疏度时也能较精确实现线阵三维SAR稀疏成像；将BCS算法应用于线阵三维SAR复数域稀疏成像,分析了BCS算法成像性能；针对BCS算法多个参数选择困难问题,基于目标散射系数指数分布、贝叶斯准则和最大后验估计原理,提出一种基于迭代最小化稀疏贝叶斯重构的线阵三维SAR稀疏成像算法,结合目标散射系数稀疏度估计、自适应参数选择和共轭梯度方法提高了算法稀疏重构性能；利用三维成像空间中目标稀疏特性,将目标位置和散射系数幅度分离进行估计,提出了一种基于稀疏目标区域预测的线阵三维SAR快速稀疏成像算法,通过粗估计稀疏目标位置减少测量矩阵维数,大大减少线阵三维SAR稀疏成像的运算量；另外,结合地基线阵三维SAR实验验证系统和外场实测数据验证了线阵三维SAR稀疏成像技术和稀疏成像方法的有效性。3.研究了相位误差情况下线阵三维SAR自聚焦稀疏成像方法。分析了不同相位误差对线阵三维SAR稀疏重构成像的影响,建立了不同维向线阵三维SAR相位误差的线性测量模型,将线阵三维SAR相位误差估计和稀疏成像转变为等幅约束线性方程的最优化求解；分析了基于相位误差估计模型松弛的最大似然估计自聚焦算法,比较了特征值松弛方法和半正定松弛方法在线阵三维SAR自聚焦稀疏成像的性能；针对稀疏欠采样情况下的线阵三维SAR回波数据,利用相位误差模型先验分布和贝叶斯准则,提出了一种基于迭代最小化贝叶斯稀疏重构的线阵三维SAR稀疏自聚焦方法,将存在相位误差的线阵三维SAR稀疏成像分解为三个线性最优化求解问题,并利用迭代逼近估计最优稀疏目标系数和误差相位,通过仿真和实测数据验证了算法的有效性。4.研究了线阵三维SAR稀疏成像中线阵阵元分布优化方法。通过理论推导分析了线阵三维SAR测量矩阵相干系数与系统模糊函数的关系,讨论了非均匀等间隔稀疏和随机稀疏阵元分布对线阵三维SAR测量矩阵相干系数的影响。基于测量矩阵相干系数最小化对线阵三维SAR稀疏成像中线阵阵元分布进行优化设计,提出了基于最小积分旁瓣比的非均匀等间隔稀疏线阵优化方法以及基于最小方差的随机稀疏线阵优化方法,通过仿真数据验证了分布优化方法的有效性。总之,本文建立了线阵三维SAR稀疏成像技术的基本原理,并在线阵三维SAR稀疏成像方法和阵列优化等方面取得了一系列有价值的研究成果,为新型线阵三维SAR稀疏成像技术研究和应用提供了重要的理论指导和技术支持。更多还原

【Abstract】 As a novel three-dimensional radar imaging technology, linear array three-dimensional synthetic aperture radar (SAR) has great and important value in military and civilian fields, such as high accuracy mapping, earth resources investigation, disasters and environmental monitoring, reconnaissance and surveillance, early warning, etc.. Limited by the traditional Nyquist sampling theorem and the classical signal processing theory, there exist some problems in the application of LASAR3-D imaging currently, including the low resolution, the high sampling ratio, the difficult of system implementation and the large number of echoes storage, transmission and processing, etc. However, as a new signal processing theorem in recent years, compressed sensing breaks the limits of the classical Nyquist sampling theorem. It can recover a sparse signal exactly with the sampled number far lower than that of the Nyquist ratio, and so has great potential in reducing the radar system sampling and improving the quality of radar imaging. Based on the compressed sensing sparse signal processing theorem, this dissertation focuses on the basic imaging principle and method research for the high resolution3-D LASAR sparse imaging, the key problems mainly including LASAR echoes linear representation, sparse reconstruction method, phase errors correction and array antenna distribution optimization, etc. The main works and innovation points are as follow:1. Research on the basic principle of LASAR sparse imaging technology. Exploiting the relationship between the LASAR echoes and the imaging space, the linear measurement models of echo signal in range direction, array plane(azimuth-cross-track plane) and the whole3-D image space are constructed respectively, and then LASAR imaging can be converted into a problem where solving the optimal resolution of the given linear equations. These linear models also provide a new idea for LASAR imaging. Further, combined the space sparsity of scatterers, a novel sparse imaging method based on compressed sensing theorem is proposed for LASAR. In addition, the linear sparse representation of scattering coefficients, the sparse sampling of echoes and the resolution of LASAR sparse imaging are discussed. For the large scale data in LASAR, a separable imaging method on range and array plane is proposed for3-D LASAR sparse imaging. Last, the performance of the classical matched filter method, the least square method and the CS sparse recovery method is analyzed through theoretical deducing.2. Research on sparse reconstruction algorithms for LASAR sparse imaging. First, the classical OMP algorithm is applied to LASAR complex data sparse imaging. For the unknown scatterers sparsity in LASAR imaging, a OMP modified algorithm, named hard threshold OMP (HTOMP) is proposed. By employing the ratio of scattering coefficient change as the iteration stop condition, HTOMP can obtain3-D LASAR image without the scatterer sparsity. Second, the promising BCS algorithm is used for LASAR complex data sparse imaging.in order to reduce the difficult of parameters selection in BCS, base on the exponential distribution of the scattering coefficient, Bayesian theory and maximum likelihood estimation, a sparse Bayesian recovery via iterative minimum (SBRIM) algorithm is proposed for LASAR sparse imaging, wherein, the sparsity estimation method, the adaptive parameter selection method and gradient conjugate method are used to improve the sparse recovery performance. Lastly, combined with the space sparsity of the scatterers in3-D imaging space, a fast sparse recovery method via target location prediction is proposed for LASAR sparse imaging. The effectiveness of LASAR sparse imaging technology and the spare imaging method is verified by some numeral simulation data and the real data obtained ground-based LASAR experimental system.3. Research on LASAR auto focus sparse imaging algorithm. First, the linear measurement models of LASAR echo signal with phase error for different direction are set up, and the phase error estimation in LASAR can be converted into solving solutions of constrain modulus quadratic program. The effect of the different types of phase errors is discussed. Based on the model relaxation and maximum likelihood estimation, the performances of LASAR sparse autofocus imaging with Eigen-value relaxation and semi-definite relaxation are analyzed. For the under-sampled LASAR echo signal, a novel sparse autofocus Bayesian recovery via iterative minimum algorithm is proposed, wherein, the LASAR autofocus sparse imaging with phase errors can be divided into three steps to finding the optimal solution of the linear equations, and the iterative estimation method is used to obtain the optimal scattering coefficients and the phase error estimation. All algorithms are performed by simulated and real experimental data.4. Research on the linear array antenna distribution optimization for LASAR sparse imaging. The relationship between the measurement matrix coherence and the LASAR system ambiguity function is studied through theoretical derivations. The effects of the uniform sparse linear array, non-uniform sparse linear array and random sparse linear array for the LASAR measurement matrix are discussed. Based on the minimum measurement matrix coherence, a distribution optimization method based on the minimum peak and the sidelobe ratio is proposed for non-uniform sparse linear array, and a distribution optimization method based on the minimum variance is proposed for random sparse linear array. Simulation results demonstrate the effectiveness of the both methods.In a word, this dissertation builds the basic principles of LASAR sparse imaging technology, and obtains a series of valuable research results for LASAR sparse imaging algorithm and linear array distribution optimization. The research results provide an important theoretical guidance and technical support for LASAR sparse imaging technology.更多还原