基于极化与多种纹理特征的高分辨率全极化SAR海面风场反演研究

项目名称： 基于极化与多种纹理特征的高分辨率全极化SAR海面风场反演研究

项目编号： No.41306192

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 天文学、地球科学

项目作者： 郑罡

作者单位： 国家海洋局第二海洋研究所

项目金额： 25万元

中文摘要： 随着高分辨率全极化合成孔径雷达（SAR）的发展，SAR能提供更多的极化和纹理信息供海面风场反演。但相应的海面风场反演研究却尚处起步阶段。本研究针对"如何利用极化和多种纹理特征来反演海面风场"这一问题进行探索，拟从以下两方面展开。1）在建立风浪全极化散射数值模型的基础上，分析极化参数与海面风场之间的关系。通过拟合模型－遗传优化和人工神经网络两种途径，探索全极化SAR地球物理模式函数的构建方法。2）采用灰度共生矩阵法，提取相关、角二阶矩及对比度等纹理特征以确定风向，在风向上进一步提取其它纹理特征，探索纹理特征与风速之间的关系，形成基于多种纹理特征的风场反演方法。另外,对传统局部梯度法进行改进，提高其效率和精度，并将其结合到上面方法中。本研究侧重基础性，将完善极化和纹理特征与海面风场之间关系的研究，并为SAR海面风场反演开拓新的思路，新颖且具有重要科学意义。

中文关键词： 全极化合成孔径雷达；海面风场；纹理特征；极化特征；

英文摘要： As the development of high resolution fully polarimetric synthetic aperture radar (SAR), it can offer more polarimetric and texture information for retrieving sea surface wind field. However, the corresponding research on sea surface wind retrieval is still at the preliminary stage. In this study, the problem of "how to use polarimetric and multiple texture features to retrieve sea surface wind field" will be explored. The study will be carried out from the following two aspects. 1) Based on the numerical model of fully polarimetric scattering from wind waves, the relationship between polarimetric parameters and sea surface wind field will be analyzed. The construction method of the fully polarimetric SAR geophysical model function will be searched by fitting model-genetic optimization and artificial neural network. 2) By using the gray level co-occurrence matrix method, the texture features (correlation, angular second moment, contrast, etc.) will be extracted to retrieve wind direction. Further, the other texture features will be extracted in wind direction. The relationship between them and wind speed will be explored. A method based on multiple texture features will be established for sea surface wind retrieval. Additionally, the traditional local gradient method will be improved for efficiency and accuracy,

英文关键词： full polarimetric synthetic aperture radar；sea surface wind field；texture feature；polarimetric feature；