Sentinel-2 is the second satellite of the world's environmental and safety monitoring system ‘Copernicus plan’, and it is an important data source for future remote sensing applications with high temporal-spatial resolution image. The simplified model for atmospheric correction (SMAC), 6S model and Sen2cor method are used to carry out atmospheric correction for Sentinel-2 satellite imagery. The upper atmospheric apparent reflectance is converted to surface reflectance, and analysis combining with measured spectral data of ground objects is carried out. After the atmospheric correction of Sentinel-2 satellite image, the spectral curves of the image and measured objects have the same change tendency with a high fitting degree. The atmospheric correction results of three models have strong correlation and high precision. The accuracy of Sen2cor method is the highest, whose determination coefficient (R2) is 0.8196, and root-mean-square error (Ermse) is 0.0388, followed by 6S model and SMAC. From the analysis of normalized differential vegetation index (NDVI), we find that NDVI values calculated by SMAC have the highest correlation with measured values, whose R2 is 0.6389, and Ermse is 0.093, followed by 6S model and Sen2cor method. Results show that the atmospheric correction accuracy of three methods is high. When the sentinel-2 satellite imagery is corrected, the image quality is improved obviously, and the availability is increased.