基于压缩感知的稀疏采样与成像方法研究

发布时间：2018-07-28 18:06

【摘要】：作为一种信息获取与处理的新理论,压缩感知已成为当前信号处理领域的一个热门研究方向。压缩感知理论指出:在满足“信号可压缩”与“观测系统与表示系统非相关”这两个条件下,能够从信号的少量采样数据中高概率复原信号,使得信号在空、时、谱等上的超分辨成为可能。由于大部分无线电信号具有可压缩性,即在某个正交/过完备字典下的编码系数是稀疏的,因此压缩感知在无线通信和成像等诸多应用中有着广泛的应用前景。例如在合成孔径雷达成像中,雷达接收到的回波可以看作是多个强散射中心回波的叠加,这种稀疏性先验使得基于压缩感知理论的稀疏成像成为可能。目前,虽然压缩感知在雷达成像显示出初步的成功,但是仍存在几个问题:第一,目前已有的压缩感知雷达成像方法基于压缩感知理论,仅利用了目标的稀疏性先验进行较少方位向脉冲下的超分辨成像。然而,随着方位向脉冲数的减少,成像质量随之迅速下降。第二,由于距离维的采样容易降低目标能量,目前已有的压缩感知雷达成像大部分均在方位维进行采样。然而,随着宽带/超宽带微波成像在安全检测与非破坏性控制等领域中的迫切需求,距离-方位维联合的超分辨技术已成为亟待解决的研究难题。针对上述问题,本文研究了基于压缩感知的稀疏采样与成像方法。具体工作如下:(1)设计了一种基于压缩感知的距离-方位联合稀疏雷达成像方法。首先对合成孔径雷达回波信号的稀疏性进行分析,研究了稀疏基的构造,实现快时间和慢时间两个维度上的联合欠采样。将该方法分别用于SAR和ISAR的超分辨成像,实验结果表明:压缩感知成像方法相对于传统微波成像方法,可以在低脉冲数下可以获得更低的旁瓣和更高的成像质量。(2)设计了一种基于显著性先验和加权L1优化的稀疏成像方法。除了目标的稀疏性先验之外,目标的显著性与几何结构可以作为先验信息以改善较少采样下的成像质量。首先利用低分辨成像结果提取视觉显著图,从中区分出显著目标区域。其次,在重构过程中对目标和背景加以不同的权值,来达到抑制背景中的杂波,同时增强目标区域中的强散射点的目标。将该方法用于256个方位维脉冲数的Yak-42数据的超分辨成像,实验结果表明:基于显著性先验的加权L1优化可以区别地对待目标和背景,实现成像时增强目标散射点同时抑制背景杂波。(3)设计了一种基于图Laplacian正则的合作式稀疏成像方法。除了目标的稀疏性先验和目标显著性之外,目标邻近距离单元的相关性也可以进一步改善成像质量。在基于显著图的加权L1优化成像基础上,挖掘临近距离单元的相似性,构成图Laplacian正则项,对原本的稀疏优化问题增加结构性约束。针对该优化问题,设计了一种基于增广拉格朗日乘子法的交替优化求解算法。将该方法用于点信号的256个方位维脉冲数的Yak-42超分辨成像,实验结果表明:图Laplacian正则项有效地降低了背景杂波中孤立的散射点,而目标上的强散射之点之间由于位于邻近距离单元,受结构性约束的影响很小。(4)设计了一种模拟信号稀疏采样的模拟信息转换器(Analog-to-Information Converter,AIC)实现。研究了基于压缩感知的模拟信号采样,设计一种基于MWC结构的硬件仿真平台,针对无线电通信系统中存在的多宽带信号,分析了MWC系统的结构、原理,实验验证其重构效果和稳定性,为稀疏雷达成像中的距离维稀疏采样的硬件实现奠定基础。
[Abstract]:As a new theory of information acquisition and processing, compressed sensing has become a hot research direction in the field of signal processing. The theory of compressed sensing indicates that, under the two conditions of satisfying "signal compressibility" and "non correlation of observation system and representation system", it is possible to recover signals from a small amount of sampled data from the signal, Because most of the radio signals are compressible, that is, the coding coefficients in a certain orthogonal / overcomplete dictionary are sparse, so compressed sensing has a wide application prospect in many applications such as wireless communication and imaging. For example, in synthetic aperture radar imaging, thunder The received echo can be regarded as the superposition of multiple strong scattering center echoes. This sparsity prior makes the sparse imaging based on compressed sensing theory possible. At present, although compression perception has shown a preliminary success in radar imaging, there are still several problems: first, the existing compression sensing radar imaging method is present. Based on the theory of compressed sensing, only using the sparsity of the target to carry out the super-resolution imaging with less azimuth to the pulse. However, with the reduction of the number of azimuth pulses, the imaging quality drops rapidly. Second, since the sampling of the distance dimension is easy to reduce the target energy, most of the existing compressed sensing radar imaging are in the azimuth. However, with the urgent need of broadband / ultra wideband microwave imaging in the fields of security detection and non destructive control, the range azimuth combination superresolution technology has become a difficult problem to be solved. In this paper, this paper studies the sparse sampling and imaging method based on the compression sensitivity. The specific work is as follows (1) a range azimuth joint sparse radar imaging method based on compressed sensing is designed. Firstly, the sparsity of the SAR echo signal is analyzed, the construction of the sparse base is studied, and the joint undersampling on two dimensions of fast and slow time is realized. The method is applied to the super-resolution imaging of SAR and ISAR respectively. The results show that the compression sensing imaging method can obtain lower sidelobe and higher imaging quality under the low pulse number compared with the traditional microwave imaging method. (2) a sparse imaging method based on significant prior and weighted L1 optimization is designed. Besides the sparsity prior of the target, the significance and geometric structure of the target can be obtained. As a priori information, the image quality under less sampling is improved. Firstly, the visual significant image is extracted from the result of low resolution imaging, and the significant target area is separated from the middle area. Secondly, the target and the background are different weights in the reconstruction process to suppress the clutter in the background and enhance the target of the strong scattering point in the target area. This method is applied to the super-resolution imaging of Yak-42 data of 256 azimuth pulse numbers. The experimental results show that the weighted L1 optimization based on significant prior can treat the target and the background discriminately, and the target scattering points are enhanced and the background clutter can be suppressed simultaneously. (3) a cooperative sparse imaging based on the graph Laplacian regularization is designed. Methods. In addition to the sparse prior and target significance of the target, the correlation of the target proximity unit can further improve the imaging quality. On the basis of the weighted L1 optimization imaging based on the saliency graph, the similarity of the adjacent distance units is excavated, and the Laplacian regular term is formed, which increases the structure of the original sparse optimization problem. In order to solve this problem, an alternating optimization algorithm based on the augmented Lagrange multiplier method is designed. This method is applied to the Yak-42 super-resolution imaging of 256 azimuth pulse numbers of point signals. The experimental results show that the figure Laplacian canonical term effectively reduces the isolated scattering points in the background clutter and the strong scattering on the target. Due to the proximity unit, the influence of structural constraints is very small. (4) a simulated signal sparse sampling analog information converter (Analog-to-Information Converter, AIC) is designed. The analog signal sampling based on compressed sensing is studied, and a hardware simulation platform based on MWC structure is designed for wireless communication. The multi wideband signal in the electrical communication system is analyzed. The structure and principle of the MWC system are analyzed. The effect and stability of the reconfiguration are verified by experiments. It lays the foundation for the hardware realization of the sparse radar imaging in the sparse sampling.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN957.52

【相似文献】