Quantitative analysis of individual sugars and acids in orange juices by near-infrared spectroscopy of dry extract

The combination of near-infrared spectroscopy (near-IR) and multivariate calibration for determination of glucose, fructose, sucrose, and citric and malic acids in orange juices was investigated. The concentrations of these components analyzed by enzymatic assays were considered as references relative to near-IR spectroscopy. Dry extract spectra of 218 orange juice samples were recorded in transmission mode between 1100 and 2500 nm. The original near-IR spectral data could be improved by mathematical pretreatments such as derivative transformations or multiplicative signal correction. Stepwise multiple linear regression (SMLR) and partial least-squares regression (PLSR) were used to create calibration models relating chemical reference values to spectral data. The prediction ability of calibration models is acceptable in comparison with the reference methods. The calibration and validation results provided by PLS-1 calibration models are slightly better than those obtained with SMLR calibration models.