Abstract: The extraction of crop phenology is very important. In recent years, there are many researches on phenology derivation from multi-temporal remote sensing data. But most of them focus on vegetation types like forests and grasslands. Since the temporal characteristics of winter wheat spectral index is different from above biomes, satisfactory phenology extraction results cannot be achieved using these methods. A novel curve fitting method is presented in this research. Using this method, detailed winter wheat phenology can be extracted from time series of SPOT/VEGETATION S10 product successfully. It was applied to a typical wheat site in Jining, Shandong Province. The results indicate that this method is robust to the noise in remote sensing dataset. Compared with other methods, time series of remote sensing data can be fitted to this model effectively, and realistic phenology results are detected.