The hyperspectral reflectances of the canopies of two varieties of rice were measured by a ASD FieldSpec Pro FR^TM in field under 3 nitrogen support levels in different stages after heading. The leaf area indices LAI corresponding to the hyperspectral reflectance, and the ultimate theoriacl yield and real yield were determined. The spectral difference vegetation index DVI, ratio of vegetation index RVI, normalized difference vegetation index NDVI and greenness G were constructed by the data of MSS wavebands, TM wavebands, SPOT wavebands, CCD wavebands on China-Brazil Resources NO.1, IKONOS wavebands, MODIS wavebands and hyperspectral wavebands. The yield estimation models of rice were established by simulating the method of remote sensing. The correlation between the yields and the hyperspectral data and vegetation indices of rice and the accuracies of yield estimation were analyzed. The yield estimation result by DVI was the best in the yield estimation models of single variable, and the results at mature stage were better than those at heading and milking stage. The results of multiplicate yield estimation model of multi-stages were better than that of single stage. They were the same results, the accuracies above 91%, for all DVI yield estimation models simulating all above wavebands for rice at mature stage, and the accuracy of MSS wavebands was the best.