Abstract:

With its quick, simple, accurate and nondestructive characteristic, NIRS (Near Infrared Transmittance Spectroscopy) is a new method for quality analysis of wheat. In this paper, spectral data were dealt with moving window average method and mahalanobis distance analysis firstly, on the basis of which, the partial least squares model of wheat grain protein was established. The experiment results indicated that the R, MSE and Er were 0.9809, 0.1130 and 1.973% respectively. In contrast with the calibration model built with the original data, the predicted results improved significantly.