To explore the spectral standardization method of NIR spectrometers with different spectral principles， the crude protein of wheat flour was taken as the specific detection standard， and the spectra of wheat flour samples collected by three NIR spectrometers with different principles were taken as the research object. The standardization method of near infrared spectrum is studied by using direct standardization(DS) and piecewise direct standardization (PDS) and simple linear regression direct standardization (SLRDS) algorithms respectively， by Euclidean distance (D)， spectral standardized error rate (SSER) spectral differences between the spectral analysis of the main， from the machine， The prediction effect was evaluated by predicting correlation coefficient (Rp)， standard deviation of prediction (RMSEP) and relative standard deviation (RPD). The results show that: 1) After the spectral standardization of the three algorithms， the difference between the master and the slave spectrum is significantly reduced， and the prediction effect of the transferred model is greatly improved. The prediction effect of the slave S2 is better than that of the slave S1. 2) Among the three algorithms， the prediction error rates of the two slave machines are the smallest after the DS algorithm is standardized. SSERave and SSERmax of the slave machine S1 are 0.9057 and 3.3667， Rp， RMSEP and RPD are 0.8949， 0.7052 and 2.2408， SSERave and SSERmax of the slave machine S2 are 0.6595 and 4.3691， Rp， RMSEP and RPD are 0.9687， 0.4105 and 4.0284， respectively. The spectral standardization of the three algorithms can be applied to different types of NIR analyzers， and the NIR correction model of crude protein content in wheat flour can be shared.