基于压缩感知的极化SAR图像分类方法研究

发布时间：2018-01-19 02:38

本文关键词： 极化合成孔径雷达图像分类极化目标分解压缩感知支持矢量机特征提取　出处：《中国民航大学》2014年硕士论文　论文类型：学位论文

【摘要】：极化合成孔径雷达(Polarimetric Synthetic Aperture Radar,PolSAR)是一种多参数、多通道雷达成像系统,它通过测量不同极化收发组合下的极化散射矩阵来反映目标的散射特性,能获得比传统的单极化合成孔径雷达更加丰富的地物信息。因而在军事、农业、水利等领域得到了广泛应用,并发挥着重要作用。极化SAR图像分类是极化SAR图像解译的重要环节,对后续的极化SAR目标检测和识别有着至关重要的作用。而特征提取是极化SAR分类的基础,特征的质量直接影响着分类结果的优劣,因此本文从特征提取的角度出发研究极化SAR图像分类方法。本文首先分析了经典的最大似然分类方法、H/?/Wishart分类方法以及基于支持矢量机的分类方法;然后介绍了压缩感知的相关理论;最后在此基础上给出了基于压缩感知和支持矢量机的非监督分类方法。该方法首先利用H/?/Wishart方法选取训练样本,解决了分类过程中训练样本难以选取的问题;然后利用压缩感知理论从原始极化相干矩阵中提取特征;最后利用支持矢量机进行特征的选择和分类。分别采用AIRSAR系统在Flevoland地区、UAVSAR系统在Hayward地区采集的实测数据进行实验,结果表明该方法具有较好的分类结果。
[Abstract]:Polarimetric Synthetic Aperture Radarr (PolSAR) is a multi-parameter. Multi-channel radar imaging system, which measures the polarimetric scattering matrix under different polarization transceiver combinations to reflect the scattering characteristics of the target. It can obtain more information of ground objects than the traditional single-polarization synthetic aperture radar, so it has been widely used in military, agriculture, water conservancy and other fields. Polarimetric SAR image classification is an important part of polarimetric SAR image interpretation. Feature extraction is the basis of polarimetric SAR classification, and the quality of feature directly affects the classification results. Therefore, this paper studies the polarimetric SAR image classification method from the angle of feature extraction. Firstly, this paper analyzes the classical maximum likelihood classification method. R Wishart classification method and support vector machine based classification method; Then the theory of compressed perception is introduced. Finally, an unsupervised classification method based on compressed sensing and support vector machine is presented. The method of / Wishart is used to select training samples, which solves the problem that training samples are difficult to select in the process of classification. Then the feature is extracted from the original polarimetric coherence matrix by the theory of compressed sensing. Finally, support vector machine (SVM) is used to select and classify the features. AIRSAR system is used in the Flevoland region. The experimental results of the measured data collected by UAVSAR system in Hayward area show that the method has good classification results.
【学位授予单位】：中国民航大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN957.52

【参考文献】