Abstract: In this study,in order to accurately extract desertification information based on remote sensing data,Spectral Mixture Analysis(SMA) was conducted in a typical area of Mu Us Sandland in semiarid region by taking farmland,sands,psammophytic vegetation,water and salinized land as endmembers.A comparison was made among SMA,TC transformation and supervised classification.The accuracy on the result was validated based on field survey data and compared with NDVI method.The result suggested that SMA could be used for extracting desertification information with an obviously better output based on remote sensing data than NDVI method.