摘要
中国首次上天的中等分辨率成像光谱仪 (CMODIS)数据具有 34谱段 ,包含了从可见光、近红外、短波红外至热红外波段非常丰富的波谱信息。但是CMODIS波段间相关性很高 ,而且数据量很大 ,若用传统的基于数理统计的分类方法 (如监督和非监督分类方法 )提取信息会有一定的困难。光谱角度制图法(SAM)是通过计算 1个测试光谱 (像元光谱 )与 1个参考光谱之间的“角度”来确定它们两者之间的相似性 ,从而达到分类目的。这一方法更符合高光谱数据的特点。文章对 2 0 0 2年 5月 2 2日取得的前 2 0个波段的CMODIS数数据分别采用监督分类方法和光谱角度制图 (SAM)法提取信息。试验结果表明采用SAM得到了比监督分类方法更好的分类效果和更多的信息。
As the first Chinese medium-resolution imaging spectromemter in SZ-3 spacecraft, the data of CMODIS contains abundant spectral information with wavelength ranging from 0 4 to 12 5 But the high correlation and the greatness of CMODIS data bring on some difficulty in information extraction of CMODIS using traditional classification method SAM is a classification method of defining the similarity between the target and the reference by calculating the 'angle' between them This paper does the test of information extraction of the CMODIS data obtained on Match 22 in 2002 using traditional supervision classification and SAM The result shows that SAM can achieve better classification result and extract more information
出处
《航天返回与遥感》
2003年第2期5-9,共5页
Spacecraft Recovery & Remote Sensing
