Search

Abstract: Near-infrared (NIR) diffuse reflectance spectroscopy has become an attractive technique for quality assurance of fruits, drugs, vegetables, and foods. In this work, the information on chemical composition of Siraitia grosvenoriiwas acquired by NIR, and then a calibration model for the prediction of total sugar content (TSC) was developed using chemometric techniques. One hundred and fifty samples were used as calibration set to develop model, while 40 unknown samples were used for prediction, using phenol-concentrated sulphuric acid assay as reference value. Several spectra preprocessing methods and wavelength selection techniques combined with partial least squares regression were compared, and then standard normal variate coupled with Savitzky–Golay derivative was selected for spectra processing and competitive adaptive reweighted sampling method was applied to extract the informative wavelengths related to the prediction of TSC. The performance of the obtained model has been evaluated by external validation, the square of correlation coefficient (R2) are 0.9097 for a separate blind set consists of 40 unknown samples. Additionally, the result of paired ttest (p?=?0.3608) illustrated that there is no significant differences between NIR and wet chemistry analysis. These results demonstrated that NIR spectroscopy has the potential to be a time saving and cost-effective method for the determination of the TSC in S. grosvenorii. Graphic abstract: