Land Cover Classification in Heihe River Basin with Time Series - MODIS NDVI Data

Normalized difference vegetation index (NDVI) is a very important vegetation index, which has been widely applied in research regarding global environmental and climatic change. In this work, 16-Day L3 Global 1 km SIN Grid NDVI data sets in Heihe River Basin from MODIS vegetation index (VI) products (MOD13A2) during 2003-2005 are extracted and used for generating a one-year new NDVI data based on a simple three-point smoothing technique which can generally capture the annual feature of vegetation change. Then we obtain the independent component images by performing independent component analysis (ICA) transform on the smoothing NDVI data as a feature extractor. Then a support vector machine (SVM) is utilized to construct classifiers based on the ICA-extracted new features for land cover classification and a land cover map of Heihe river basin was obtained. At last, the accuracy assessment results prove that the classification framework proposed in this paper is efficient.