Crop information extraction in China based on NDVI characteristic curve

Cropland is one of the most important types of land surface. It provides the resources and environment on which human being relies for existence. This paper uses MODIS surface classification data and NDVI remote sensing data as data source to calculate the NDVI time series characteristic curve of cropland by histograms. We further obtain the 2001-2011 cropland distribution under 500m resolution with decision tree classification. Finally, the paper uses 2009 MERIS Globcover data as reference and calculates the producer´s accuracy. The results show that: the NDVI time series characteristic curve based on histograms is more suitable for extracting cropland information. The extraction producer´s accuracy shows an increase compared with cropland producer´s accuracy of MODIS.