Application of Fourier Transform Near-Infrared (FT-NIR) and Fourier Transform Infrared (FT-IR) Spectroscopy Coupled with Wavelength Selection for Fast Discrimination of Similar Color of Tuber Flours

Abstract

This research aimed at providing a fast and accurate method in discriminating tuber flours having similar color by using Fourier transform near-infrared (FT-NIR) and Fourier transform infrared (FT-IR) spectroscopy in order to minimize misclassification if using human eye or avoid adulteration. Reflectance spectra of three types of tubers (consisted of Canna edulis, modified cassava, and white sweet potato) were collected to develop a multivariate model of partial least-squares discriminant analysis (PLS-DA). Several spectra preprocessing methods were applied to obtain the best calibration and prediction model, while variable importance in the projection (VIP) wavelength selection method was used to reduce variables in developing the model. The PLS-DA model achieved 100% accuracy in predicting all types of flours, both for FT-NIR and FT-IR. The model was also able to discriminate all flours with coefficient of determination (R2) of 0.99 and a standard error of prediction (SEP) of 0.03% by using 1st Savitzky Golay (SG) derivative method for the FT-NIR data, as well as R2 of 0.99 and SEP of 0.08% by using 1st Savitzky Golay (SG) derivative method for the FT-IR data. By applying the VIP method, the variables were reduced from 1738 to 608 variables with R2 of 0.99 and SEP of 0.09% for FT IR and from 1557 to 385 variables with R2 of 0.99 and SEP of 0.05% for FT NIR.