一种用于遥感成像系统的压缩感知编码矩阵设计

发布时间：2018-06-11 17:31

本文选题：压缩感知 + 编码矩阵　；参考：《航天返回与遥感》2017年05期

【摘要】：压缩感知作为突破传统奈奎斯特定理限制的一种信号处理的新途径,近年来受到了诸多研究领域的广泛关注,特别是在遥感成像方面。该理论中,编码矩阵的设计起着非常关键的作用。事实上,原始信号能否被有效压缩,接收端能否将原始信号精确重构,都依赖于编码矩阵设计的优劣。然而,目前常见的编码矩阵普遍不利于硬件的实现,尤其是遥感成像中的采样更是要求计算简单、省电的设备,所以这是影响压缩感知在遥感成像领域推广的主要障碍之一。文章的研究目的是找到一种新的编码矩阵,既有良好的压缩感知采样性能,又有利于针对遥感应用的硬件实现和降低硬件成本。鉴于分块压缩感知在重构时的优势即提升重构速度和品质,提出了基于分块的二级尺度编码矩阵设计,即在第一次分块的基础上再次分块,并以此基础设置编码矩阵。通过实验模拟实际的硬件采样过程,分析所设计编码矩阵的不足之处,并对其进一步优化,使得所搭建的硬件成像平台中对图像采样的数据可以在终端进行高品质的重构。
[Abstract]:As a new way of signal processing which is restricted by the traditional Nyquist theorem, compressed sensing has received extensive attention in many fields of research in recent years, especially in remote sensing imaging. In this theory, the design of the coding matrix plays a very important role. In fact, whether the original signal can be effectively compressed and whether the receiver will be able to be the original or not The accurate reconstruction of the initial signal depends on the design of the coding matrix. However, the common coding matrix is generally not conducive to the implementation of the hardware. Especially, the sampling in remote sensing imaging is more important for computing simple and electricity saving devices. Therefore, this is one of the main obstacles that influence the promotion of compressed sensing in remote sensing image field. In order to find a new coding matrix, it not only has good compressed sensing sampling performance, but also is beneficial to hardware implementation and hardware cost reduction for remote sensing applications. In view of the superiority of the block compression perception in the reconstruction of the reconstruction speed and quality, a block based two level scale coding matrix is designed, that is, in the first block. On the basis of this, the coding matrix is set up again and the coding matrix is set on this basis. By simulating the actual hardware sampling process, the shortcomings of the designed coding matrix are analyzed, and the further optimization is made, which makes the data of the image sampled in the hardware imaging platform can be reconstructed with high quality at the final end.
【作者单位】：南京理工大学计算机学院;
【基金】：国家自然科学基金(61273251)
【分类号】：TN911.73

【相似文献】