基于压缩感知的层析SAR成像方法研究

发布时间：2018-03-06 10:48

本文选题：压缩感知　切入点：层析SAR成像　出处：《北京建筑大学》2017年硕士论文　论文类型：学位论文

【摘要】：合成孔径雷达(SAR)三维成像技术不仅具有传统SAR系统所具备的全天时、全天候等优势,而且解决了三维成像中将高度向投影到二维平面时产生的叠掩问题,其能将目标高度向与距离向完全分离,实现目标的三维成像。在地形测绘、军事侦察、资源勘测、灾害的精准评估以及森林资源与生态监测的动态监测等方面具有广泛的应用前景,是成像雷达科学工作研究者的研究热点。层析SAR三维成像利用多次航过或者多天线单次航过获取的同一地区SAR二维图像,在距离垂向上运用层析原理进行孔径合成,从而实现高分辨三维成像。合成孔径雷达三维成像无须针对飞行器进行飞行轨迹的刻意控制,并且无须进行特定的位置控制,就能具备真正意义上的三维成像能力,在成像雷达中具有很好的应用前景。但是,层析SAR三维成像研究不够成熟,从理论研究到应用实现仍需要有很长的发展过程。最近几年由于压缩感知理论的提出,解决了数据获取中必须要满足奈奎斯特采样定理而所带来的数据存储量大和传输困难的问题。该理论指出,当信号具有稀疏性时,通过求解非线性最优化问题,可以用远低于奈奎斯特采样定理对信号进行重构。层析SAR三维成像理论中,对目标的观测数量远低于距离垂向上目标(即信号)的点数,但由于距离垂向上的目标(信号)点数又满足了稀疏性的特点,因此基于压缩感知架构的层析SAR三维成像理论应运而生。本文基于此,针对压缩感知的层析SAR三维成像开展了深入研究。本文首先研究了层析SAR成像的几何模型、数学模型;针对层析SAR成像的数学模型,着重研究基于压缩感知的层析SAR三维成像方法。以压缩感知作为基础理论,分别研究了正交匹配追踪、正则化正交匹配追踪、压缩感知匹配追踪、广义正交匹配追踪等层析SAR成像算法;在研究和分析算法的基础上,运用仿真数据和真实数据对这些方法进行验证。仿真数据反映了算法有效性。根据层析SAR成像的流程,论文还研究了SAR图像配准方法,并根据真实数据对其进行验证,得到良好的配准效果。针对真实数据的验证,论文选取了北京地区鸟巢附近的盘古七星酒店作为研究对象,采用TerraSAR-X数据进行试验,真实数据的试验结果表明论文方法的合理性。本文的创新点在于:(1)提出基于相关函数的图像精配准方法。该方法首先对图像粗配准,然后进行分块插值处理,提高了图像的配准精度。(2)采用正交匹配追踪、正则化正交匹配追踪、压缩感知匹配、广义正交匹配等算法进行压缩感知的仿真,对其重构结果进行对比分析,评价了上述四种算法的优劣。
[Abstract]:Synthetic Aperture Radar (SAR) 3D imaging technology not only has the advantages of all-day, all-weather and so on as the traditional SAR system, but also solves the problem of overlay when the height is projected to the two-dimensional plane in 3D imaging. It can completely separate the height and distance of the target and realize the three-dimensional imaging of the target. In topographic mapping, military reconnaissance, resource survey, The accurate assessment of disasters and the dynamic monitoring of forest resources and ecology have wide application prospects. Tomographic SAR 3D imaging is a hot research topic in imaging radar science. Using SAR 2D images of the same area obtained by multiple voyages or single voyage with multiple antennas, the tomography principle is used to synthesize aperture at vertical distance. In order to achieve high resolution 3D imaging, synthetic Aperture Radar (SAR) 3D imaging does not require deliberate control of the flight trajectory of the aircraft, nor does it require specific position control, in order to have a true three-dimensional imaging capability. It has a good application prospect in imaging radar. However, the study of tomography SAR 3D imaging is not mature enough, and it still needs a long development process from the theoretical research to the application. In recent years, the compression sensing theory has been proposed. The problem of large data storage and difficult transmission caused by Nyquist sampling theorem in data acquisition is solved. The theory points out that when the signal is sparse, the nonlinear optimization problem is solved. The signal can be reconstructed with a far lower sampling theorem than Nyquist. In the theory of SAR tomography, the number of observations on a target is much lower than the number of points in the vertical target (the signal). However, because the number of targets (signals) in vertical distance satisfies the characteristics of sparsity, the theory of three-dimensional tomography (SAR) imaging based on compression sensing architecture emerges as the times require. In this paper, the geometric model and mathematical model of tomography SAR imaging are studied firstly, and the mathematical model of tomographic SAR imaging is also studied. Based on the theory of compressed sensing, orthogonal matching tracking, regularized orthogonal matching tracking and compressed perceptual matching tracking are studied respectively. Based on the research and analysis of the algorithms, these methods are verified by simulation data and real data. The simulation data reflect the validity of the algorithm. According to the process of tomographic SAR imaging, This paper also studies the SAR image registration method, and verifies it according to the real data, and obtains good registration effect. In view of the real data verification, the paper selects the Pangu Seven Star Hotel near the Bird's Nest in Beijing as the research object. The experimental results of TerraSAR-X data show that the method is reasonable. The innovation of this paper lies in: (1) an image registration method based on correlation function is proposed. Then the block interpolation processing is carried out to improve the registration accuracy of the image. The algorithm of orthogonal matching tracking, regularized orthogonal matching tracking, compressed perceptual matching, generalized orthogonal matching and so on is used to simulate the compression perception. The reconstruction results are compared and analyzed, and the advantages and disadvantages of the above four algorithms are evaluated.
【学位授予单位】：北京建筑大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN957.52

【参考文献】