Use of Visible and Short-Wavelength Near-Infrared Spectroscopy and Least-Squares Support Vector Machines for Non-Destructive Rice Wine Quality Determination

H. Y. Yu,X. Y. Niu,Y. B. Ying,X. X. Pan

DOI: https://doi.org/10.13031/2013.36236

2011-01-01

Transactions of the ASABE

Abstract:The feasibility of Vis/SW-NIR spectroscopy for non-destructive determination of rice wine composition in sealed, colored bottles was investigated. The Vis/SW-NIR spectra were collected with a fiber spectrometer system (500-1200 nm) in sealed, colored bottles. Statistical equations were established between the reference data and Vis/SW-NIR spectra by partial least squares regression (PLSR) and least-squares support vector machines (LS-SVM), respectively. The LS-SVM models outperformed PLSR models, with higher correlation coefficients for validation (r(val)) for alcohol content, titratable acidity, and pH of 0.960, 0.942, and 0.866, together with root mean square error of validation (RMSEP) values of 0.115% (v v(-1)), 0.071 g L-1, and 0.021, respectively. Based on the results, it was concluded that the Vis/SW-NIR spectrometer system was suitable for non-destructive rice wine quality determination in colored bottles, and LS-SVM was a reliable multivariate method for NIR spectroscopy analysis. This study would be helpful for a future real-time system for rice wine quality determination.