Prediction of protein content of intact wheat seeds with near infrared reflectance spectroscopy (NIRS)

In the present article, the measurement of intact wheat seeds' protein content with near infrared reflentance spectroscopy(NIRS) was studied. The wavelength range of 1 100-2 498 nm. was used. The distinguishability of NIRS machine was 2 nm. Firstly the representative wheat samples with different protein contents were selected and the original spectra of wheat were obtained using NIRS machine. Then through scatter correction and maths treatment, spectrum noise were decreased. Finally regression methods used multiple linear regression, principal component regression and modified partial least squares respectively. The result showed that after optimizing all the factors, the best calibration model of equation was chosen using "first derivative" +"Standard Multiplicative Scatter Correctiong SMSC" + "Modified Partial Least Squares, MPLS". RSQ, SECV and 1-VR of the obtained calibration model were 0.94, 0.42 and 0.87 respectively. A set of representative individual wheat samples were uesed to check the model, and prediction coefficient of determination was 0.88. Protein content of wheat could be preidicted quickly and scathelessly by using the NIRS measurement. It was feasible to be used in early generation selection in wheat quality breeding process.