Rapid Near Infrared Spectroscopic Screening of Chemical Parameters in Semi-hard Cheese Using Chemometrics

Abstract

A remote reflectance module measured the near infrared spectra (400 to 2500 nm) of 107 samples of semi-hard Danbo cheese. The spectral measurements were correlated with the contents of dry matter, fat, and total nitrogen in the cheese using the chemometric method of partial least squares regression. On selected sample subsets, correlations were made with soluble nitrogen, amino acid nitrogen, and ammonia to examine the relationships between the degree of maturation and near infrared spectra. The best results for fat (r = 0.99; root mean standard error of prediction = 0.52), and for total nitrogen (r = 0.98; root mean standard error of prediction = 0.06) were obtained when the spectra were pretransformed with multiplicative signal correction. The best result for dry matter (r = 0.98; root mean standard error of prediction = 0.58) was obtained when spectra were pretransformed by dividing each wavelength by its standard deviation, indicating that the optimal pretransformation of near infrared spectra depends on which cheese component is being modeled. Correlations with degree of maturation showed that near infrared spectra cannot replace traditional analyses. The degree of maturation did not interfere with the modeling of fat, nitrogen, or dry matter.