基于特征值和Singh分解的全极化Radarsat-2图像分类
发布时间:2018-01-22 05:09
本文关键词: 全极化SAR图像分类 Singh分解 Cloude分解 极化总功率 SVM 出处:《地理空间信息》2016年05期 论文类型:期刊论文
【摘要】:基于模型的分解发展较快,但存在负功率、体散射过估计、未充分利用相干矩阵等问题,考虑到基于模型分解的优点,采用Singh分解提取极化信息,同时用散射角、极化熵和极化总功率进行补充,再利用SVM对山东禹城地区全极化Radarsat-2数据进行分类。为验证该方法的有效性,将其与H/α/A-Wishart和Yamaguchi-SVM两种分类方法进行比较。结果表明,该方法分类效果较好,总体精度分别提高了6.4%和3.48%。
[Abstract]:The decomposition based on model has developed rapidly, but there are some problems such as negative power, over estimation of volume scattering, and insufficient use of coherent matrix. Considering the advantages of model-based decomposition, Singh decomposition is used to extract polarization information. At the same time, the scattering angle, polarization entropy and total polarization power are used to supplement, and then SVM is used to classify the fully polarized Radarsat-2 data in Yucheng area, Shandong Province. It was compared with two classification methods, H- 伪 / A-Wishart and Yamaguchi-SVM, and the results showed that the classification effect of this method was better than that of H- 伪 / A-Wishart and Yamaguchi-SVM. The overall accuracy was improved by 6.4% and 3.48 respectively.
【作者单位】: 南京大学地理与海洋科学学院;江苏省地理信息技术重点实验室;江苏省地理信息资源开发与利用协同创新中心;济南市环境监测中心站;78125部队;
【基金】:国家重点基础研究发展计划资助项目(2010CB951503) 中国科学院战略性先导科技专项资助项目(XDA05050106)
【分类号】:P237
【正文快照】: 非相干极化分解是全极化SAR图像分类的重要方法之一,近年来得到广泛应用[1-6]。非相干极化分解主要有基于特征值分解和基于模型分解两大类[7-13]。基于模型的分解得到几种确定的散射机制,但是这是一种理想化的处理,实际情况常常包含未知的散射机制。近年来,基于模型的分解有较
【相似文献】
相关期刊论文 前1条
1 ;Index for Volume5,2008[J];Journal of Mountain Science;2008年04期
,本文编号:1453769
本文链接:https://www.wllwen.com/kejilunwen/dizhicehuilunwen/1453769.html