论文部分内容阅读
多光谱遥感影像因具有丰富的波谱信息,提高了地表覆盖的辨识能力,利用遥感数据高精度自动提取专题信息是目前研究的热点和难点。本文以北京市ASTER影像为例,通过对城市生态环境中土地类型及其光谱特征规律分析,组合归一化差异植被指数、修正归一化差异水体指数和归一化差异建筑指数三种指数,制作组合指数新影像。对组合指数影像采用基于支持向量机的面向对象分类方法进行农业用地信息提取,同时将该方法分别与基于原始影像、组合指数影像的最大似然及支持向量机的分类方法进行对比分析。实验结果表明:组合归一化差异指数影像压缩了数据维数,降低了覆盖地物相关性,易于农业用地信息提取。对组合指数影像采用基于支持向量机的面向对象分类方法精度达95.701%。
Multi-spectral remote sensing image has rich spectrum information, which improves the recognition ability of surface coverage. It is a hot and difficult issue to use the remote sensing data to automatically extract thematic information with high accuracy. In this paper, taking Beijing ASTER image as an example, this paper analyzes the land types and their spectral characteristics in the urban ecological environment, and combines the normalized difference vegetation index, the normalized difference water index and the normalized difference construction index , Making a new composite index image. The method of object-oriented classification based on support vector machine is used to extract agricultural land information for composite index image, and the method is compared with the classification method based on maximum likelihood of original image and composite index image and support vector machine. The experimental results show that the combination of normalized difference index images reduces the dimensionality of data, reduces the correlation of covered features and facilitates the extraction of agricultural land use information. The object-oriented classification method based on SVM has a precision of 95.701% for composite index images.