Abstract: In order to resolve the fault of traditional methods，MODIS-NDVI time series data, land-use plan in 1985 and Agricultural Statistics data of Hubei province were used to obtain MODIS-NDVI time series curve from April 2007 to March 2008 in Hubei Province. As a result of impacts by clouds, atmosphere during transmission, remote sensing images’ curve fall short of statistical data, even sometimes appeared apparent dislocation phenomena with phenophase. In order to accurately reflect the true features of land surface, we used the Fourier transform to make time series data to frequency domain from space domain, used Savitzky-Golay filter for denoising processing and analysis, and then compared the results with that of Hanning denoising filter in ERDAS. Results showed that the former is better. It not only can remove noise, non-circular stripes, but also retain maximum details of the characteristics. Now, adopted the inverse Fourier transform to the results of Savitzky-Golay to get relatively smooth curve. Combined with phenological characteristics of Hubei Province, we made analysis on curve, and then according to the specific information curve, made classification to the original images from December 2007 to February 2008 in Hubei province. Two types of winter fallow fields and non-winter fallow fields were got considering the results of 9 image classification results, integrated the total results of winter fallow field’s distribution, area and number of days in Hubei province. The total areas of winter fallow fields is 5564.3 kilometers and most of them located in the southeast of Hubei province.