Abstract: Winter wheat has been one of the main food crops in the North China Plain. Crop phenology refers to the specific events of the crop life cycle that change with the seasons. It is of great significance to accurately obtain the regional phenological periods of winter wheat in agricultural production. Traditional monitoring phenological periods can rely primarily on laborious field observations, indicating the limited time and space. Remote sensing technology can be expected to monitor long-term and large-scale phenology. However, there are single data sources of traditional phenology monitoring technologies by remote sensing, leading to medium and low spatial resolution. The low accuracy of extraction cannot fully meet the demands of agricultural production. Normalized Difference Vegetation Index (NDVI) time series data is currently the main data source to extract the phenological periods through remote sensing. In this study, the multi-source remote sensing images were utilized to calculate and reconstruct the long-time series NDVI data. The optical remote sensing images were taken from GF-1, HJ-2, Landsat-8 and Sentinel-2. The spatial distribution maps of winter wheat were extracted in the phenological periods at field scale over six years. The smooth curves were then selected to truly reflect the wheat growing state in the natural state. The extraction model was also established for the phenology periods of winter wheat. Firstly, a temporally uneven NDVI image dataset was constructed using multi-source optical remote sensing data. Then a uniform daily NDVI dataset was obtained using interpolation methods. The NDVI time series curves were reconstructed through Savitzky-Golay (SG) filtering, in order to eliminate the noises of time series data. According to the relationship between the characteristics of winter wheat NDVI curves and phenological features, the raster images were finally extracted for three phenological stages: start of season (sowing), peak of season (heading), and end of season (post-harvesting) using the extremum method. The raster images for the maturity stage were extracted using the dynamic threshold. The pixel-scale phenological data was converted to the field scale using computer vision vectorization. Compared with the field survey data, the dates of the start of the season and the peak of the season by remote sensing were consistent with the sowing and heading dates from field data, respectively. The field harvesting dates were between the maturity and the end dates of the season by remote season. The results show that the sowing period for winter wheat in the 2017-2018 season was later than in the other five years; The heading period in the 2019-2020 and 2020-2021 seasons was significantly earlier than in the other four years; The maturity period in the 2019-2020 season was earlier than in the other five years, and the end period in this season was also earlier than in the other five years. Compared with the multi-year climate data and phenological extraction, if the temperature was higher, precipitation was sufficient and sunshine duration was longer in the period of overwintering and regreening, while the heading period was advanced accordingly. Therefore, the interannual difference in temperature, precipitation, and sunshine duration can lead to the interannual variation in the winter wheat heading period. The findings can provide scientific evidence to monitor the growth and development of winter wheat.