The timely and effective prediction of crop nitrogen content can provide reasonable information support for monitoring and diagnosis of crop nitrogen nutrients. By collecting spectral information of flowering and filling stages of winter wheat and corresponding plant nitrogen content, the correlation between the original spectrum, reciprocal spectrum, logarithmic spectrum, differential spectrum, reciprocal logarithmic spectrum, logarithmic differential spectrum, reciprocal differential spectrum and plant nitrogen content was analyzed. Sensitive bands under different spectral reflectance conditions were screened. Multivariate linear regression was used. Stepwise regression and partial least squares（PLS） were used to establish remote sensing estimation models of plant nitrogen content. The results showed that:（1） The optimum sensitive bands at flowering stage were FR₆₈₉, FLR₇₅₃, FR₇₆₈, F/R₇₄₉and FLR₄₉₅,while the optimum sensitive bands at filling stage were 1/R955, LR₉₅₅, L/R₉₅₅, F/R₄₉₀and R₉₅₅.（2） The most effective model was the PLS for flowering stage, and R² and RMSE for modeling and validation were 0.72, 0.12 and 0.62,0.16, respectively.（3） P was used at filling stage. The model established by LS has the best effect. R~2and RMSE of modeling and validation are 0.70, 0.11, 0.32 and 0.18, respectively. The study found that the remote sensing estimation model of plant nitrogen content at flowering stage had high accuracy and reliability, which provided theoretical basis for better detection of nitrogen use efficiency and precise fertilization of winter wheat.