基于线性光谱混合模型分解MODIS多光谱影像的研究

《水土保持研究》[ISSN:1005-3409/CN:61-1272/P] 卷: 18 期数: 2011年03期页码: 10-14 栏目: 出版日期: 2011-06-20

Title:: Classification of MODIS Multispectral Images Based on Line Spectral Mixture Model

作者:: 陆广勇^1,3, 杨勤科^2,1, 张少伟^1,3, 王海江^1,3; 1. 中国科学院水利部水土保持研究所, 陕西杨凌 712100;
2. 西北大学, 西安 710127;
3. 中国科学院研究生院, 北京 100039

Author(s):: LU Guang-yong^1,3, YANG Qin-ke^2,1, ZHANG Shao-wei^1,3, WANG Hai-jiang^1,3; 1. Institute of Soil and Water Conservation, Chinese Academy of Sciences and Ministry of Water Resources, Yangling, Shaanxi 712100, China;
2. Northwest University, Xi’an 710069, China;
3. Graduate University of Chinese Academy of Sciences, Beijing 100039, China

摘要:: 为了获取多时相的土地覆盖基础数据以支持区域土壤侵蚀定量评价，基于线性光谱混合模型分解MODIS多光谱影像，并对分解结果进行了定性、定量评价。结果表明，结合像元年内植被指数变化特征，基于线性混合像元分解，可解译出耕地、林地、草地、裸地、水体、居民地等类型。分类结果与2006年TM分类结果的总体一致性为64.46%，Kappa系数为0.519 9，土地覆盖类型分类结果可靠；各类端元估算误差基本小于20%，且与对应TM分类结果具有相关性，总体精度较好；林地端元能够较好地反映植被盖度信息。基于LSMM分解MODIS影像可为区域环境研究提供可靠的土地覆盖类型图和植被覆盖信息。

Abstract:: Soil erosion monitoring works require multi-temporal land cover data on a regional scale.This paper estimated the abundance of land cover components using MODIS09GA multispectral images based on linear spectral mixture model(LSMM) and assessed the accuracy.The results showed the overall accuracy of the final land cover map was 64.46%,and the Kappa coefficient was 0.519 9;the estimation error of the endmember fraction is less than 20% for most of land cover components.There is fine linear correlation between the abundance of land cover components and the classification of Landsat image.For the forest endmember,the linear correlation coefficient is 0.79.It is suggested that,for MODIS image,applying LSMM in land cover classification is feasible and is promising to extracting vegetation fraction.

参考文献/References:

更新日期/Last Update: 1900-01-01