Rapid determination of reducing sugar content in sweet potatoes using NIR spectra

Reducing sugar plays a vital role in edible quality and processing properties of sweet potatoes. This study was aimed to quantitatively predict the reducing sugar content of sweet potatoes through mining near-infrared (NIR) reflectance, Kubelka-Munk (KM) and absorbance spectra in the 900–1700 nm range, respectively, by using partial least squares (PLS) algorithm. Stepwise regression combined with the regression coefficient (SRRC) method was applied to select optimal wavelengths to optimize original full band PLS models. It was found that SRRC-KM-PLS model built with 14 optimal wavelengths selected from KM spectra had better performance with a regression coefficient of prediction (rP) of 0.952 and root mean square error of prediction (RMSEP) of 0.264 g/100 g. Two-sample F-test and t-test (P > 0.05) results indicated the statistical soundness and predictive validity. In conclusion, it is reasonable and feasible to detect reducing sugar content in sweet potatoes via NIR spectra in a rapid way.