线阵三维SAR稀疏成像方法研究

发布时间：2018-06-23 04:35

本文选题：线阵三维SAR + 压缩感知　；参考：《电子科技大学》2017年硕士论文

【摘要】：作为一种新型的三维SAR成像模式,线阵三维SAR不仅克服了传统二维SAR成像的阴影遮挡效应、空间模糊、顶底倒置等缺点,而且相对于圆周SAR、层析SAR等三维SAR模式,具有运动轨迹容易控制的优点,在军事和民用方面都有着广泛的应用。由于受Nyquist采样定理的约束以及传统基于匹配滤波的SAR成像算法分辨率的限制,线阵三维SAR成像存在着回波数据量巨大、成像分辨率低等缺点。压缩感知理论指出,如果信号是稀疏的或可压缩的,那么可以用低于Nyquist采样定理规定的采样率恢复原信号。在三维SAR成像中,成像的目标场景往往具有稀疏性,因而压缩感知理论在提高SAR成像分辨率、降低SAR回波数据量上有着重大的应用。本文将压缩感知理论运用于线阵三维SAR中,主要研究了基于稀疏重构的线阵三维SAR成像方法,具体研究内容和创新点如下:1、简述了线阵三维SAR成像基础与压缩感知基本理论。首先,介绍了线阵三维SAR的几何模型,给出了其回波信号模型。其次,在模糊函数的基础上分析了传统成像算法的分辨率限制,并且介绍了两种传统的SAR成像算法及优缺点。最后,介绍了压缩感知的基本理论、分析了线阵三维SAR的线性观测模型,为后面将压缩感知理论运用于线阵三维SAR高分辨率成像提供了理论依据。2、研究了基于贪婪算法的SAR成像方法。首先,阐述了正交匹配追踪(OMP)算法的流程。OMP算法因具有流程简单、运算效率较高、重构误差较小等特点得到了广泛的应用。其次,总结了近年来针对OMP算法相关学者提出的一些改进算法。这些算法主要从元素选择的准则、系数更新、候选集数目等方面进行改进。最后,重点研究了梯度追踪(GP)算法,该算法相对于OMP算法具有运算效率高、空间存储低的优点。3、提出了一种基于门限的梯度追踪(TBGP)算法。GP算法需要预设稀疏度,而对于SAR成像来说,很难获得场景的真实稀疏度。针对此问题,运用对比度和变化率代替预设稀疏度作为迭代次数的准则,提出了TBGP算法。该算法不仅保留了GP算法在运算时间和空间存储上的优势,又克服了GP算法需要预设稀疏度的缺点。仿真及实测数据实验验证了上述结论。4、研究了迭代最小稀疏贝叶斯重构(SBRIM)算法。分析了SBRIM算法的原理以及步骤流程。作为稀疏贝叶斯一类算法,相对于其他类算法来说,具有灵活性强、重构精度高的特点。仿真实验结果表明SBRIM算法相对于传统BP算法具有高分辨的成像能力。5、提出了加权迭代最小稀疏贝叶斯重构(WSBRIM)算法。与SBRIM算法不同,WSBRIM算法是对代价函数中的L1范数向进行加权约束,并且采用了不同的函数近似L1范数,减小了运算量。详细阐述了WSBRIM算法的原理以及步骤。对所提到的相关算法做了仿真实验分析和性能比较。仿真实验结果表明,在重构精度上,WSBRIM算法的重构误差最小,重构精度最好;而在运算效率上,本文所提的TBGP算法更具有优势。最后,三维SAR实测数据成像结果验证了WSBRIM算法在实际场景中的有效性。
[Abstract]:As a new model of three-dimensional SAR imaging, linear array 3D SAR not only overcomes the shortcomings of the shadow shielding effect of the traditional two-dimensional SAR imaging, the space blurred, the top bottom inversion, but also has the advantages of easy control of the motion trajectory relative to the circumference SAR, the tomography SAR and other three-dimensional SAR modes, and has a wide application in military and civil applications. Under the constraints of the Nyquist sampling theorem and the limitation of the resolution of the traditional SAR imaging algorithm based on the matched filtering, the linear array 3D SAR imaging has a large amount of echo data and low imaging resolution. The compression perception theory indicates that if the signal is sparse or compressible, it can be sampled with a sample lower than the Nyquist sampling theorem. In 3D SAR imaging, the target scene of the imaging often has sparsity, so the compression perception theory has a great application in improving the resolution of SAR imaging and reducing the amount of SAR echo data. In this paper, the compression perception theory is applied to the linear array 3D SAR, and the three-dimensional SAR imaging based on sparse reconstruction is mainly studied. The main research contents and innovation points are as follows: 1, the basic theory of linear array 3D SAR imaging and compression perception are briefly described. First, the geometric model of linear array 3D SAR is introduced, and the echo signal model is given. Secondly, the resolution limitation of the traditional imaging algorithm is analyzed on the basis of fuzzy function, and two traditional S are introduced. AR imaging algorithm and its advantages and disadvantages. Finally, the basic theory of compressed sensing is introduced, the linear observation model of linear array 3D SAR is analyzed. In order to provide a theoretical basis for the application of compressed sensing theory to 3D SAR high resolution imaging of linear array, the SAR imaging method based on greedy algorithm is studied. First, the orthogonal matching tracking (OMP) is expounded. The algorithm's process.OMP algorithm has been widely used because of its simple process, high operation efficiency and less reconstruction error. Secondly, some improved algorithms proposed by the relevant scholars in recent years are summarized. These algorithms are mainly improved from the criteria of element selection, the updating of the coefficients, the number of candidate sets and so on. Finally, the algorithm is improved. This paper focuses on the gradient tracking (GP) algorithm. Compared with the OMP algorithm, the algorithm has the advantage of high computing efficiency and low spatial storage,.3. A gradient tracking (TBGP) algorithm based on threshold (TBGP) algorithm.GP algorithm needs presupposition sparsity, but for SAR imaging, it is difficult to obtain the real sparsity of the scene. By replacing the preset sparsity as the criterion of the number of iterations, the TBGP algorithm is proposed. The algorithm not only preserves the advantages of the GP algorithm in the operation time and space storage, but also overcomes the shortcoming of the default sparsity in the GP algorithm. The simulation and the measured data experiments verify the above conclusion, and study the iterative minimum sparse Bayesian reconstruction (SBRI). M) algorithm. The principle and step process of SBRIM algorithm are analyzed. As a sparse Bias class algorithm, compared with other class algorithms, the algorithm has the characteristics of strong flexibility and high reconstruction precision. The simulation results show that the SBRIM algorithm has high resolution image ability.5 compared with the traditional BP algorithm, and the weighted iterative minimum sparse Bias is proposed. The WSBRIM algorithm is different from the SBRIM algorithm. The WSBRIM algorithm is a weighted constraint on the L1 norm in the cost function, and the different functions are used to approximate the L1 norm, and the computation is reduced. The principle and steps of the WSBRIM algorithm are elaborated in detail. The simulation experiment analysis and performance comparison of the related arithmetic are made. The results show that the WSBRIM algorithm has the least reconstruction error and the best reconstruction precision in the reconstruction precision, and the TBGP algorithm proposed in this paper is more advantageous in the operation efficiency. Finally, the results of three-dimensional SAR measured data show the effectiveness of the WSBRIM algorithm in the actual scene.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN957.52

【相似文献】