文献类型：期刊文献

中文题名：基于多角度高光谱CHRIS影像的隆宝滩湿地遥感分类方法研究

英文题名：Study on Remote Sensing Classification Method of Long Baotan Wetland Based on CHRIS/PROBA

作者：韦玮[1] 李增元[1] 谭炳香[1] 徐海生[1]

机构：[1]中国林业科学研究院资源信息研究所

年份：2011

卷号：24

期号：2

起止页码：159-164

中文期刊名：林业科学研究

外文期刊名：Forest Research

收录：CSTPCD;;Scopus;北大核心:【北大核心2008】；CSCD:【CSCD2011_2012】；

基金：中央级公益性科研院所基本科研业务费专项资金"(IFRIT200906)"

语种：中文

中文关键词：遥感;高光谱;多角度;穗帽变换;湿地;监督分类

外文关键词：remote sensing; hyperspectral; multi-angle; tasseled cap; wetland; supervised classification

分类号：S771.8

摘要：采用青海省隆宝滩地区的多角度高光谱CHRIS遥感数据,通过研究+36°、0°和-36°三个角度影像的组合变换,提出影像变换+不同角度波段组合的方法,用以获取地物的分类信息。该方法首先对0°影像进行穗帽变换,选择其湿度图像,再与+36°和-36°影像的第4波段(0.461μm)进行RGB组合,生成新的彩色合成影像,然后再进行支持向量机(SVM)的监督分类。结果显示,利用该方法对隆宝滩湿地分类的精度可达到90.02%;而利用传统的监督分类对0°影像直接进行分类,其精度为75.46%。由此可见,利用不同角度信息进行波段组合的方法,大大提高了高光谱影像进行湿地信息提取的精度,为湿地信息提取提供了一个有效的方法。
This paper provides a new method of improving the classification accuracy of wetland in Long Baotan area in Qinghai Province,by studying the image transformation and band combinations of three angle images of ＋36°,0° and-36°,which were devived from the multi-angle CHRIS hyperspectral remote sensing data.Firstly,the tasseled cap transformation was used to the 0° CHRIS image.Secondly,a new color composite image of RGB was generated by combining the humidity image of 0° with the 4-band（0.461μm） of ＋36° and-36° images,and then,the Support Vector Machine,SVM,a supervised classification method was carried out on the new RGB image.The studies showed that the classification accuracy of the new combination method in different angle images of CHRIS approached to 90.02%,which was greatly improved then 75.46% of traditional supervised classification accuracy,and it also provide an effective method to extract wetlands information.