For the purpose of automatically obtaining a large area of winter wheat planting area, phenological information in medium spatial resolution remote sensing images based on time series curves is usually used to identify and extract. However, in actual engineering projects, if phenological information is used only, the accuracy is low. Therefore, a method based on time series curve data classification model and image segmentation is proposed for winter wheat identification. Firstly, the normalized difference vegetation index (NDVI) time series curve of multi-source data is constructed, and the NDVI time series data are smoothed and denoised by harmonic analysis of time series (HANTS) method. Then, via coordinate transformation of NDVI time series, three parameters of band mean, standard deviation and square mean are obtained to construct a new classification model so as to improve the difference between winter wheat and other crops; finally, by combining segmentation results of spatial resolution data, spatial structure information of the image is used to improve accuracy of feature boundary. Taking Jiangning District of Nanjing as an example, the authors used 21 multi-source images of GF-1, Landsat8 and Sentinel-2A from December 2017 to June 2018, and the final extraction accuracy reached 98.74%, which is better than results of other methods. This method provides agricultural management departments with accurate geographic information data on planting area and distribution of winter wheat.