摘要
利用NDVITs空间进行全国土地覆盖分类的方法。该方法利用1995年NOAA10天合成的ch4、ch5通道亮温,先计算出陆地表面温度(Ts),然后用最大值合成法计算每月的最大Ts和NDVI,以每月最大Ts和NDVI建立NDVITs空间。根据像素点(NDVI,Ts)在空间中的位置矢量,求出矢量在空间中的方向角度,并作归一化处理,得到温度植被角度(NTVA)。对12个月NTVA做主成分变换提取前三个主分量,辅以全年总NDVI和大于0℃Ts积温,用模糊K—均值法进行全国土地覆盖分类。研究结果表明,基于NDVITs空间的NTVA与NDVI、Ts一起作为分类特征在土地覆盖分类中具有较高的分类精度,能够取得较好的分类效果。
In this paper, a method based on NDVI-Ts space is proposed for Chinese land cover classification. The ten days composite ch4 and ch5 light temperatures for 1995 were used to estimate land surface temperature with split window method. Then the monthly land surface temperature and NDVI were produced with the maximum value composite from the 10-day composite Ts and NDVI. With the monthly Ts and NDVI, the monthly NDVI-Ts spaces were created. And pixel vector direction in the NDVI-Ts space was described with NTVA (Normalized Temperature Vegetation Angel). Principal Components Analysis (PCA) was used to compress the 12 monthly NTVA images and three Principal Components were extracted. Fuzzy K-mean algorithm in clustering was used with the three Principal Components, summation of monthly NDVI and accumulated >0℃ Ts image to produce China land cover classification map. The result revealed that remarkable improvement for land cover classification can be reached with the combination of NDVI and land surface temperature..
出处
《遥感学报》
EI
CSCD
北大核心
2005年第1期93-99,共7页
NATIONAL REMOTE SENSING BULLETIN
基金
国家高技术研究发展计划863-308-13-03(02)
国家高技术研究发展计划2003AA131020
中国科学院知识创新工程重大项目(KZCX1SW0102)。