Cereal Chem 63:151-154   |  VIEW ARTICLE
Discrimination of Bread-Baking Quality of Wheats According to Their Variety by Near-Infrared Reflectance Spectroscopy.

M. F. Devaux, D. Bertrand, and G. Martin. Copyright 1986 by the American Association of Cereal Chemists, Inc. 

Assessment of the bread-baking quality of French soft wheat by near-infrared reflectance (NIR) spectroscopy, using standard mathematical treatments and reference tests, has had limited success. New mathematical treatments of NIR spectra are reported that assign a technological value to wheat varieties to differentiate potential bread-baking quality categories. The seven most cultivated soft wheat varieties in France were chosen for this study. Two preliminary mathematical treatments were applied to reduce the number of spectral data (351 wavelengths) without loss of information. These were fast Fourier transform and principal component analysis of the Fourier coefficients. By these methods, the log (1/R) spectral values were transformed to 10 independent synthetic variables that were subsequently treated by multiple discriminant analysis. The first mathematical treatment was made to discriminate and recognize the seven varieties, and 74% of unknown samples were identified (79 samples for calibration and 42 for prediction). The two varieties that were principally confused (Fidel and Talent) were of the same bread-baking quality. The second mathematical treatment concerned three bread-baking categories: good, unsuitable, and irregular. Correct category assignments were made for 95% of unknown samples. Advantages of those mathematical methods are discussed.