基于局部与非局部策略的极化SAR相干斑抑制研究

发布时间：2018-10-16 21:36

【摘要】：极化合成孔径雷达(POLSAR)是一种多参数、多通道的微波成像雷达,对地物具有一定的投射性,能够获取不同极化状态下目标的散射回波,更好的描述目标所包含的信息,已成为遥感领域的一种新技术,是合成孔径雷达(SAR)发展的新趋势,被广泛应用于军事、国防、环境、农业等各个领域。但是由于相干成像的原理性缺陷,极化SAR系统获取的目标信息与实际的信息之间存在一定的偏差,导致极化SAR数据中存在相干斑噪声,严重的影响了数据的解译和后续应用,因此相干斑噪声的抑制成为极化SAR数据应用中的首要问题。本文介绍了SAR数据以及极化SAR数据的统计分布和相干斑噪声模型,提出了三种相干斑抑制算法,主要内容如下:1.在已有极化SAR数据双边滤波的基础上,借鉴基于相对全变差的边缘检测方法,提出了一种基于边缘检测和双边滤波的极化SAR相干斑抑制算法,检测出图像的匀质区域和边缘纹理细节区域,在不同的区域对双边滤波的亮度信息平滑参数设置不同的值,使其更适合所处理的像素。该算法简单快速,可以直接对极化协方差矩阵C或极化相干矩阵T进行处理。实验结果表明,该算法不仅对相干斑噪声抑制具有相对不错的性能而且对边缘纹理细节信息的保持也相对较好。2.结合双边滤波和非局部均值滤波,提出了一种基于非局部双边滤波的极化SAR相干斑抑制算法,通过非局部的思想选择尽可能多的相似图像块,并对这些相似图像块采取双边滤波,最后对相似图像块双边滤波的结果进行非局部加权平均,既考虑了图像块的结构相似性又考虑了单像素的像素值相似性。此算法结合了局部和非局部策略,不但考虑了双边滤波的优点也考虑了非局部均值滤波的优点。实验结果表明,不论在匀质区域还是在边缘纹理细节区域该算法对极化SAR数据图像的相干斑抑制效果都相对较好。3.将处理含有加性高斯白噪声图像的SVD算法应用到SAR数据的相干斑抑制上,针对SAR数据的统计分布以及相干斑噪声模型,在获取SVD样本矩阵时采用适合SAR数据的相似性距离代替原始的欧氏距离,并提出适应SAR数据相干斑噪声的SVD阈值参数计算公式。除此之外,充分考虑极化SAR数据的特性,将应用到SAR数据相干斑抑制上的SVD方法扩展到极化SAR数据上,在Span数据上获取相似图像块和SVD阈值参数,在保持极化信息方面有一定的作用。实验结果表明,该算法在SAR数据和极化SAR数据上的相干斑抑制均能达到较好的性能,而且更大的优势在于对单视极化SAR数据处理具有一定的有效性。本文工作得到了国家自然科学基金(No.61173092)、新世纪优秀人才支持计划(No.66ZY110)和陕西省科学技术研究发展计划项目(No.2013KJXX-64)资助。
[Abstract]:Polarimetric synthetic Aperture Radar (POLSAR) is a kind of multi-parameter, multi-channel microwave imaging radar, which can project objects to a certain extent, and can obtain scattering echo of the target in different polarization state, and describe the information contained in the target better. It has become a new technology in the field of remote sensing and a new trend in the development of synthetic Aperture Radar (SAR), which has been widely used in military, national defense, environment, agriculture and other fields. However, due to the principle defects of coherent imaging, there is a certain deviation between the target information obtained by polarized SAR system and the actual information, which leads to speckle noise in polarized SAR data, which seriously affects the interpretation and subsequent application of the data. Therefore, the suppression of speckle noise becomes the most important problem in the application of polarimetric SAR data. In this paper, the statistical distribution and speckle noise model of SAR data and polarized SAR data are introduced, and three speckle suppression algorithms are proposed. The main contents are as follows: 1. Based on the bilateral filtering of polarimetric SAR data and the relative total variation, an edge detection algorithm based on edge detection and two-sided filtering is proposed for polarimetric SAR speckle suppression. The homogeneous region and the edge texture detail area of the image are detected, and different values of the brightness information smoothing parameters of the bilateral filtering are set in different regions to make it more suitable for the processed pixels. The algorithm is simple and fast and can directly process the polarization covariance matrix C or the polarized coherence matrix T. Experimental results show that the proposed algorithm not only has relatively good performance for speckle noise suppression, but also maintains the edge texture details relatively well. 2. A polarimetric SAR speckle suppression algorithm based on non-local two-sided filtering and non-local mean filtering is proposed, which selects as many similar image blocks as possible. These similar image blocks are filtered by two sides. Finally, the results of the similar image blocks are weighted by non-local weighted averaging. The structural similarity of the image blocks and the pixel similarity of the single pixel are taken into account in both the structural similarity of the image blocks and the pixel similarity of the single pixel. This algorithm combines local and non-local strategies, not only considering the advantages of bilateral filtering, but also considering the advantages of non-local mean filtering. The experimental results show that the proposed algorithm has better speckle suppression effect on polarized SAR images both in homogeneous region and in edge texture detail region. The SVD algorithm with additive Gao Si white noise image is applied to speckle suppression of SAR data. Aiming at the statistical distribution of SAR data and the speckle noise model, In obtaining the SVD sample matrix, the Euclidean distance is replaced by the similarity distance suitable for SAR data, and a formula for calculating the threshold parameters of SVD for speckle noise of SAR data is proposed. In addition, considering the characteristics of polarized SAR data, the SVD method applied to speckle suppression of SAR data is extended to polarimetric SAR data to obtain similar image blocks and SVD threshold parameters on Span data. In maintaining polarization information has a certain role. Experimental results show that the proposed algorithm can achieve better performance in both SAR data and polarimetric SAR data, and its greater advantage lies in its effectiveness in single-view polarimetric SAR data processing. This paper is supported by the National Natural Science Foundation (No.61173092), the New Century Talent support Program (No.66ZY110) and the Shaanxi Province Scientific and technological Research and Development Program (No.2013KJXX-64).
【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

【共引文献】