This study combined spectral mixture analysis (SMA)with supervised classification for extracting the VegetationFractional Coverage (VFC)of Uxin country in Mu Us desert,China.Three periods of multi-temporal satellite imageswhich are landsat 5 TM images acquired between 1988-1991,Landsat 7 ETM+images acquired between 1999-2001and HJ-1B CCD images acquired in 2009,were respectively utilized to retrieve the normalized differential vegetationindex (NDVI)of Uxin country in different periods.Simultaneity,a supervised method of classification was employed todivide the image into three classes including vegetation,bare soil and water.Set the mean of ten highest NDVI value ofvegetation class as the pure vegetation NDVI value,and the mean of ten lowest NDVI value of bare soil class as thepure bare soil NDVI value,based on which,the dimidiate pixel model was established to extract VFC in Uxin countryin different periods.The result shows that the VCF of Uxin country had a increasing trend in the past 30 years,andmany desert land converted into vegetation covered land which is own to the construction of the ¨Three North 〃protection forest.