多源遥感影像的集值特征选择与融合分类  

Feature selection for fusion classification of multisource remotesensing images on set-valued features

作　　者：孙亮[1] 韩崇昭[1] 康欣[1] 

机构地区：[1]西安交通大学电子与信息工程学院

出　　处：《电波科学学报》2004年第4期479-484,共6页Chinese Journal of Radio Science

基　　金：国家重点基础研究发展规划项目 (2 0 0 1CB30 94 0 3)

摘　　要：提出了一种多源遥感影像集值特征选择 (SVFS)与评价新方法。该方法根据训练样本在原始特征空间中的多维统计特性建立了关于其集值特征的目标信息系统 ,对于给定的集值特征间的二元关系 ,利用属性约简进行了特征选择并给出了相应的重要度评价算法。对多源遥感影像的融合分类实验表明 。A set-valued feature selection (SVFS) method for fusion classification of multisource remote sensing images is proposed. The Framework consists of three sections: establishment of object information system from the statistical characteristic of the training samples in the original feature space, feature selection which is performed by the attribute reduction based on the binary relation about the set-valued features, and calculation of significance for each reduced attribute set. The fusion classification experiment shows that it is an effective method based on generalized rough set for the feature selection of multisource remote sensing images.

关 键 词：多源遥感影像 集值特征 特征选择 属性约筒 广义粗糙集 融合分类 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]