Studies Of EOS/MODIS Data Processing And Land Cover Remote Sensing In China

In this paper, the methods of EOS/MODIS LIB data processing are discussed based on the format and the technical reports of MODIS IB datasets, including data preprocessing, data extracting, data calibrating and projection transformation etc. Some problems in data processing are solved. A method of de_striping, EDF adjusting, is adopted. And the earth location data matching between different resolution pixels are discussed including data interpolating. A software to process MODIS data is made in VC++6.0 development environment based on the HDF4 library and the General Cartographic Transformation Package(GCTP). From which users can extract earth view data of different band or resolution and other ancillary data according to users' demand. It supports most current projection types and can transform projection coordinates forwardly or inversely and can calculate NDVI and EVI. It is useful for processing a large amount of MODIS data.For the MODIS application of land cover, using 1km and 500m MODIS IB data of year 2001 from NSMC, 16_day and 32_day NDVI product was obtained through MVC. The result shows that MVC can apparently eliminate the effects of cloud contamination and obtain relatively pure land pixels. Selecting some clear_sky MODIS data of Beijing, using multispetral images of 250m, land cover around Beijing was analyzed. An unsupervised dynamic clustering experimentation was done using ISODATA method. The result shows that the MODIS multispectral data has some capacities of distinguishing land cover types. Land cover of China region was analyzed using the 1km NDVI and the relative ancillary data. The result shows that the distribution of vegetation in china is correlative with space and the NDVI seasonal variations are different between various ecosystems. The land cover classification can be based on this character. The land cover of northeast of China was classified by Bayes maximun likelihood rule.