In the current nitrogen inversion study, there are fewer studies on the agronomic parameter of rice grain nitrogen. Since rice grain nitrogen content directly influences crop nutrition, it is crucial to conduct a study on the estimation of rice grain nitrogen based on the relationship between hyperspectral features and plant nitrogen content to assess the nutritional status and quality of the crop. For this purpose, unmanned aerial hyperspectral remote sensing images of rice at the stage of jointing, flowering and maturing were acquired , and a study on estimating rice grain nitrogen content was carried out by using plant nitrogen content as a bridge at the rice experimental base in Gaoqiao Town, Changsha County, Hunan Provincial Academy of Agricultural Sciences (HPAAS). Firstly, the fitting precision R2 between the narrow-band normalized difference vegetation index (N-NDVI) and plant nitrogen content was analyzed, and the region of extreme maximum value Ω was obtained and the center of the plant nitrogen-sensitive band was calculated. Then the optimal width of the sensitive band for rice plant nitrogen content estimation was automatically determined by constructing an internal connection rectangle. Subsequently, a statistical model linking plant nitrogen content to rice grain nitrogen was established. Finally, the estimation of rice grain nitrogen content was realized by constructing N-NDVI based on the optimal bandwidth, and the accuracy was verified. The results showed that the optimal sensitive bandwidth of N-NDVI for plant nitrogen content, which was automatically screened by using the internal rectangle, achieved high accuracy in the inversion of rice plant nitrogen content and rice grain nitrogen content in all periods. The coefficients of determination R2 between the measured and predicted values of rice grain nitrogen content ranged from 0.410 9 to 0.610 6, the normalized root mean square errors (NRMSE) ranged from 11.33% to 16.85%, and the mean relative errors (MRE) ranged from 9.53% to 13.24%. The prediction accuracy of each fertility stage from high to low was as follows: maturing stage, jointing stage and flowering stage. At the maturing stage, the N-NDVI constructed with the central wavelength of the sensitive band at 629.85/701.93 nm,corresponding to the hyperspectral optimal bandwidth of ±6 nm, had the highest accuracy in estimating plant nitrogen content and predicting the nitrogen content of rice grains (R2=0.590 0, NRMSE=14.06%, MRE=11.59%). In conclusion, the inversion method of rice grain nitrogen content proposed was feasible and can achieve accurate estimation of rice grain nitrogen content, which provided an idea for predicting rice grain nitrogen content in gramineous cereal crops.