进入高分SAR时代的舰船类型识别技术研究

项目名称： 进入高分SAR时代的舰船类型识别技术研究

项目编号： No.61471024

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 无线电电子学、电信技术

项目作者： 郎海涛

作者单位： 北京化工大学

项目金额： 86万元

中文摘要： 高分SAR使舰船类型识别成为可能，是当前研究的前沿课题。舰船类型识别的准确率主要取决于舰船特征的提取和选择，以及分类器设计两个核心问题。舰船特征提取和选择方面，现有研究提出的舰船类型对舰船类型敏感，但易受环境条件的影响。本项目提出分析舰船特征对类型和环境的敏感性的新方法，发展新的对类型敏感而对环境条件不敏感的舰船特征。在分类器设计方面，现有方法难以有效融合多元异构的舰船特征构造良好的分类器。本项目提出采用核方法融合舰船特征，发展基于稀疏核学习-支持向量机的舰船类型分类器。本项目的研究成果将形成一套系统的高分SAR舰船类型识别技术，为我国执行海上维权工作提供技术支持，为我国早日成为海上强国贡献力量。

中文关键词： 合成孔径雷达；图像解译；高分辨率；特征提取与选择；分类器

英文摘要： High resolution SAR makes it possible to identify the category of ships, which is the forefront of current research. The accuracy of ship classification depends on two key points, that is extracting and selecting optimal ship features, and designing a robust classifier. Ship features, which are proposed by existing methods, are usually sensitive to ship types, while unstable to environment. This project proposes a novel method to analysis the sensitivity of features to ship type and environmental condition, and develops a serial of features both sensitive to type and stable to environment. In the aspect of classifier design, it is difficult to fuse multiple heterogeneous features by a single classifier for existing methods. In this project, we propose to utilize kernel method to fuse heterogeneous ship features, and develop sparse kernel learning-SVM classifier. The outcome of this project will develop a systemic technology for ship classification based on high resolution SAR, and provide technical support to maritime rights protection of our country.

英文关键词： SAR;Image interpretation;High resolution;Feature extraction and selection;Classifier