Land Cover Classification in China Based on Chosen Bands of MODIS

Abstract

MODIS data with high spectral and temporal resolutions were used as input parameters for regional land cover classification in China.First,EVI,NDWI and NDSI were calculated as input spectral features on the basis of an annual time series of twelve MODIS 8-day composite reflectance images(MOD09) acquired during the year of 2007.The three indices were added to the image form a 10 spectral bands image.The authors employed the mean Jeffries-Matusita distance as a statistical separability criterion and classification accuracy of SVM to evaluate the contribution of different bands for land cover classification.Once the aim was achieved,the monthly three largest contribution spectral bands(EVI、B7 and B4) were dealt with.The Principal Component Analysis(PCA) method and its first three principal components were used as input parameters for SVM classification.The result shows that the three largest contribution spectral bands together with temporal information as input parameters can reach certain high classification accuracy(78.04%) at moderate spatial scales without other accessorial data.