| Large-scale crop type automatic identification methods were studied based on remote sensing images. The characteristics of NDVI time series of cotton in Xinjiang were analyzed, and then the threshold values of extracting planting area in different zones were set and the extraction models of planting area in different zones were established, and finally the spatial distribution of cotton of 2009 were obtained. Compared with years of average statistical data，the extrac-tion accuracy is more than 78%. Secondly， we used NDVI difference model to monitor the cotton growth condition of Xinjiang in 2009，and compared it with the average cotton growth of last five years. The results showed that there were different characteristics in the spatio-temporal distribution. |