Abstract: Abstract: In order to measure plant nitrogen content (PNC) rapidly and accurately in different growth stages, the optimal regression model for PNC was constructed based on variable importance projection - partial least squares - Akaike's information criteria (VIP-PLS-AIC) and corresponding PNC data. In this research, 16 spectral indices sensitive to nitrogen and chlorophyll were constructed by using of winter wheat canopy reflectance obtained in National Precision Agriculture Experimental Base from 2014 to 2015. The model was verified by using of data at flag leaf stage from 2012 to 2013. Results showed that in jointing stage the related degree order between VIP evaluation sixteen vegetation index and winter wheat PNC can be drawn as follows: PPR> Red_Width> SRPI> NPCI> NPQI> SIPI> Red_Area> MCARI/MTVI2> TCARI> PSNDc> MCARI> DCNI> REPGAUSS> REP> PRI> SR(533,565). In booting stage the order between VIP and PNC can be drawn as follows: PPR> SRPI> NPCI> NPQI> MCARI/MTVI2> SR(533,565)> PRI> SIPI>REPGUSS>REP>Red_Area>PSNDc>Red_ Width>DCNI>MCARI>TCARI. In anthesis stage the order between VIP and PNC can be described as PPR> NPQI> MCARI> MCARI/MTVI2> TCARI> DCNI> REPGAUSS> REP> SR(533,565)> SRPI> NPCI> PSNDc> Red_Width> PRI> Red_Area> SIPI. In filling stage, the order between VIP and PNC can be described as TCARI> MCARI> NPQI> DCNI> SIPI> MCARI/MTVI2> PPR> Red_Area> REPGAUSS> REP> PSNDc> Red_Width> SR(533,565)> PRI> SRPI> NPCI. The PNC model of winter wheat based on AIC at jointing stage using four vegetation indices as independent variables was the optimal. At flag leaf stage, flowering stage and filling stage they were five, four and six kinds, respectively. The determined coefficients (R2) and root mean square error (RMSE) during four growth stages were 0.71, 0.86, 0.75, 0.46 and 0.23%、0.13%、0.12%、0.15%, respectively. At booting stage the independent variables respectively were VPPR, VSRPI, VNPCI, VNPQI and VMCARI/MTVI2. The booting stage in 2012 to 2013 data was used to validate and the booting stage was the optimal stage for estimating winter wheat PNC using hyperspectral data. The results showed R2 and RMSE of validation set at booting stage were 0.81 and 0.41%. Besides, both prediction model and verification model had higher accuracy and reliability. The estimation result of winter wheat PNC based on coupling model VIP-PLS-AIC was ideal and provided an effective method for predicting winter wheat PNC by remote sensing. The overall results showed that the PNC of winter wheat can be reliably monitored with the canopy spectral methods established in the study.