Using 162 rice samples as materials, the method of ammonia-gas-sensing electrode was employed to determinate the protein content, while atomic absorption apectrophotometry (ASS) was employed to determinate the contents of Ca, Mg, Fe, Zn, Cu and Mn, colorimetry with phosphate-molybdenum-blue complex was employed to determinate the content of P, and flame photometry was employed to determinate the content of K in brown rice. The relationships of protein content and mineral content, different mineral element contents in brown rice were investigated. the chemometrical method of partial least squarere gression was used to establish the calibration model of protein content in brown rice. The results showed that the elemental concentrations(mg．kg-1) in brown rice are in turn of P>K>Mg>Ca>Zn>Fe>Cu>Mn. Significant positive correlations were found between protein content and mineral contents, including P, K, Cu, Mn. In addition, the optimal model was developed by the spectral data pretreatment of the first derivative in 11995.7-7498.3cm-1 and 6102-4597.7cm-1, by analyzingspectral data pretreatment and light frequency ranges. This model’s calibration coefficient and determination coefficient were 0.9886 and 0.8991, respectively. The model showed significant correlation and lower error between near-infrared value and true value. The germplasm of rice resource with high protein content and rich mineral contents, Such as Xiao Heigu, Xiao Hongmi and Zi Nuomi had been selected. Good calibration equation was successfully developed for protein content, the equation show satisfactory determination coefficients. Finally, a probably effective way to improve protein content of rice was proposed. Combination of some special characteristics, shch as protein content, P, K, Cu and Mn etc, was one of the effective approaches to increase nutrient of rice. This NIRS-assisted-selection could be a very efficient method to improve protein content and mineral contents in rice breeding programs.