The split-window algorithm was recommended to retrieve the land surface temperature of Anhui Province using MODIS image data. Based on the supervised classification methods and the NDVI methods proposed by Sobrino and Qin Zhihao, the land surface emissivity was estimated. The inversion results were compared with NASA land surface temperature products, and the average error was around 1 K. Furthermore, the results were compared with the surface temperature from the corresponding position weather station observation at the same time when satellite transits. The results reveal that they have the same trend and a high correlation, so it can reflect the land surface temperature of Anhui Province intuitively.