Abstract: Accurate monitoring of crop nitrogen status and pre-harvest prediction of grain quality are important for precision nitrogen management and quality-oriented wheat production. However, canopy hyperspectral data contain numerous highly correlated bands, and the resulting information redundancy may reduce modeling efficiency and weaken effective spectral responses to crop nitrogen status. In addition, grain protein formation is a continuous physiological process involving nitrogen accumulation, remobilization, and translocation across multiple growth stages, making information from a single growth stage insufficient to fully characterize the dynamics associated with grain protein accumulation. To address these issues, a leaf nitrogen content (LNC) inversion method based on spectral transformation and characteristic-band selection was developed, and multi-growth-stage LNC information was further used to predict grain protein content (GPC). Canopy hyperspectral reflectance data were collected at the jointing, flowering, and grain-filling stages of winter wheat. The original spectra were transformed using first- and second-derivative processing, and characteristic bands were selected using Pearson correlation analysis and the Boruta algorithm. Gaussian process regression (GPR), random forest regression (RFR), and support vector regression (SVR) were then used to establish stage-specific LNC inversion models. In parallel, two GPC prediction pathways were constructed: a direct pathway based on canopy spectral features and an indirect pathway using LNC derived from hyperspectral data as an intermediate physiological variable. For the indirect pathway, LNC information from individual growth stages and combinations of multiple growth stages was evaluated to investigate the contribution of integrated nitrogen status to GPC prediction. The results showed that first-derivative reflectance was more strongly correlated with LNC than the original spectra, with maximum absolute Pearson correlation coefficients of 0.74, 0.79, and 0.88 at the jointing, flowering, and grain-filling stages, respectively. Characteristic-band selection further improved LNC estimation compared with full-band input. After Boruta screening, the best validation coefficients of determination (R²) for LNC estimation at the three growth stages were 0.67, 0.80, and 0.77, respectively, with corresponding root mean square errors (RMSEs) of 0.38%, 0.38%, and 0.36%. These results indicated that first-derivative transformation combined with Boruta-based feature selection reduced spectral redundancy while retaining bands closely associated with variations in LNC. For direct GPC prediction, model performance varied among growth stages, and the flowering-stage GPR model achieved the highest accuracy, with a validation R² of 0.87 and an RMSE of 0.56%, indicating that canopy spectral information at flowering was closely associated with subsequent grain protein formation. For indirect GPC prediction, the model integrating LNC information from the jointing, flowering, and grain-filling stages performed better than models based on LNC from a single growth stage, achieving a validation R² of 0.75 and an RMSE of 0.62%. The improvement obtained by integrating LNC from multiple growth stages indicates that nitrogen status at different developmental phases provides complementary information related to grain protein formation. LNC at the jointing stage reflects nitrogen accumulation during vegetative growth, flowering-stage LNC represents nitrogen status around the transition from vegetative to reproductive development, and grain-filling-stage LNC is associated with nitrogen remobilization and translocation to developing grains. Integrating these growth-stage-specific nitrogen signals therefore provides a physiological linkage between canopy hyperspectral responses and the temporal processes underlying GPC formation. The resulting “canopy hyperspectral reflectance–multi-growth-stage LNC–GPC” pathway establishes a hyperspectral inversion framework for winter wheat grain protein driven by leaf nitrogen content and provides a methodological basis for jointly monitoring crop nitrogen nutrition and predicting grain quality before harvest.