摘要
利用MODIS数据所反演的每8d一景,全年共46景的时间序列叶面积指数(LAI)图像,分析江西省不同类型地物的LAI时间序列谱,并对地物进行分类。首先,利用最小噪声比变换技术(MNF)将噪声从数据中分离;然后,通过纯净像元指数(PPI)从LAI时间序列谱中提取5类主要地物类型终端单元(Endmember),从而对地物进行分类并制图;最后,结合2000年江西省兴国县1 10万比例尺的土地利用/覆盖矢量图对本研究分类结果进行检验。结果表明,该方法的地物分类精度达到74.45%,其分类方法是有效可行的。
In this paper, time-series LAI was retrieved from 46 MODIS images in Jiangxi province, and the images were obtained every 8 days from January 1, 2001 to the end of that year. The curve of time-series LAI was used to detect the land cover and use. Based on the hyperspectral analytical system, the authors used minimum noise fraction rotation(MNF) to extract noise from data, and employed pixel purity index(PPI) to extract the end-member of the chief land type. In this way, the land types of Jiangxi province could be classified and mapped. The classification result was verified by using the 1︰100 000 land use/cover vector map of Xingguo county in 2000, and the classification precision was above 74.45%. The result shows that the application of time-series chart to classification is effective.
出处
《国土资源遥感》
CSCD
2004年第3期5-7,22,共4页
Remote Sensing for Land & Resources
关键词
遥感
MODIS
LAI
地物
分类
Remote sensing
MODIS
LAI
Land types
Classification