基于散射模型的极化相似性建模及滤波应用

发布时间：2018-11-16 13:44

【摘要】：极化合成孔径雷达(POL-SAR)技术是在合成孔径雷达(SAR)技术的基础上发展来的,它是一种新型的遥感技术。它不仅继承了SAR的能全天时,全天候的工作,高分辨率等优势,还能够获得多种极化组合方式下的回波信号,给人们提供更准确、更丰富的地物目标散射信息。所以,极化SAR被广泛的应用到了民用领域和军事领域。但是,极化SAR存在的相干斑噪声不仅降低了图像的质量,还给极化SAR的地物分类,目标的检测和识别等研究领域带来了困难。现有的很多相干斑抑制算法在对图像进行降斑处理的同时不能很好的保持图像中的细节信息和散射特性。本文通过对合成的极化SAR数据和真实的极化SAR数据进行分析,提出了两种极化相似性度量方法:基于三分类的极化相似性度量方法和基于统计分布模型的极化相似性度量方法并将它们应用到极化SAR相干斑抑制中。本文的主要工作:(1)通过对吴月珍提出的极化相似性度量方法的研究,提出了一种基于三分类的极化相似性度量方法。该相似性度量方法首先对极化数据进行混合四分量(HPD)分解,得到散射特征向量,根据数据散射类型的稳定性,将数据分成三类,对每一类数据进行不同的相似性度量。将该相似性度量方法应用到极化SAR局部滤波中,实验结果表明,该相似性度量方法能够比较准确的找到像素点的相似集合。(2)通过对极化SAR数据的散射向量分析,提出了一种新的极化相似性度量方法,即基于统计分布模型的极化相似性度量方法。该相似性度量方法首先对极化SAR数据进行HPD分解,得到散射特征向量,通过对像素点散射向量和其相似像素点的散射向量在各个分量上差值的分析,得出差值都近似服从正态分布,通过统计分析,我们得出各个分量上差值的均方差,根据正态分布的性质,找到阈值,通过将差值的绝对值与阈值比较来判断两个像素是否相似。将该相似性度量方法应用到Lee滤波中,极化SAR图像滤波后的结果表明,该相似性度量方法能够准确的找到像素点的相似点。通过对NL-Lee滤波的分析,提出了一种基于散射统计模型和NL-Lee的极化SAR相干斑抑制算法,通过对合成的极化SAR数据和真实的极化SAR数据进行降斑处理,可以得出该算法取得了比pretest滤波和NL-Lee滤波更好的效果,说明该滤波方法能够很好的保持极化SAR图像的散射特性。
[Abstract]:Polarimetric synthetic Aperture Radar (POL-SAR) is a new type of remote sensing technology, which is developed on the basis of (SAR) technology. It not only inherits the advantages of SAR, such as all-day, all-weather work, high resolution and so on, but also can obtain the echo signal under various polarization combinations, which can provide more accurate and abundant scattering information of ground objects. Therefore, polarized SAR is widely used in civil and military fields. However, speckle noise in polarimetric SAR not only reduces the image quality, but also brings difficulties in the field of ground object classification, target detection and recognition of polarimetric SAR. Many existing speckle suppression algorithms can not preserve the detailed information and scattering characteristics of the image while reducing the speckle. In this paper, the synthetic polarized SAR data and the real polarized SAR data are analyzed. Two kinds of polarimetric similarity measurement methods are proposed: polarization similarity measurement based on three categories and polarization similarity measurement based on statistical distribution model and applied to polarimetric SAR speckle suppression. The main work of this paper is as follows: (1) by studying the polarization similarity measurement method proposed by Wu Yuezhen, a polarization similarity measurement method based on three categories is proposed. The similarity measurement method firstly decomposes the polarization data into four components (HPD) and obtains the scattering eigenvector. According to the stability of the scattering type the data are divided into three types and each kind of data is measured with different similarity. The similarity measurement method is applied to the polarimetric SAR local filtering. The experimental results show that the similarity measure method can find the similarity set of pixels accurately. (2) the scattering vector analysis of polarized SAR data is carried out. A new polarimetric similarity measurement method based on statistical distribution model is proposed. The similarity measure method firstly decomposes polarimetric SAR data by HPD and obtains the scattering eigenvector. The difference between the pixel scattering vector and the scattering vector of its similar pixel is analyzed. The difference is similar to normal distribution. Through statistical analysis, we get the mean variance of the difference on each component. According to the nature of normal distribution, we find the threshold value. The absolute value of the difference value is compared with the threshold value to determine whether the two pixels are similar. The similarity measurement method is applied to Lee filtering. The results of polarimetric SAR image filtering show that the similarity measurement method can accurately find the similarity points of pixels. Based on the analysis of NL-Lee filter, a speckle suppression algorithm based on scattering statistical model and NL-Lee is proposed for polarimetric SAR speckle suppression. The synthetic polarimetric SAR data and the real polarimetric SAR data are processed by speckle reduction. It can be concluded that this algorithm has better effect than pretest filter and NL-Lee filter. It shows that the proposed filtering method can keep the scattering characteristics of polarized SAR images well.
【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】：2015
【分类号】：TN957.52

【相似文献】