Abstract: In order to be fast, simple and accurate determination of the protein content of wheat, this paper puts forward modeling method of near infrared spectral analysis technology combined with BP neural network using genetic algorithm (GA). Spectral data were rationally divided by SPXY algorithm, and making use of successive projections algorithm (SPA) compressed preprocessed data and extracted the best sensitive wave points as the input of GA-BP neural network to establish the calibration model of the protein content of wheat. Root-Mean-Square Error of Prediction (RMSEP) and prediction correlation coefficient (R) of the model were 1.3379 and 0.979 and compared with the correction model of BP neural network. The result showed that: GA-BP neural network model has fast speed of convergence, short training time and high accuracy, which is able to achieve rapid and efficient detection on the protein content of wheat.