Abstract
The low-cost, ultra-compact and handheld microNIR spectrometer over the spectral range of 1150–2150 nm was explored to detect the adulteration of wheat flour in this study. Eight varieties of cassava flour were used as adulterants and were adulterated in wheat flour at five adulteration levels of 5, 10, 20, 30 and 40%. Both principal component analysis-linear discriminant analysis (PCA-LDA) and partial least squares discriminant analysis (PLS-DA) methods were employed to establish 2-class, 3-class and 6-class discriminant models, using different types of preprocessed absorbance spectra. The overall prediction accuracies of the 2-class discriminant models all achieved over 95.00% in separating the pure and adulterated wheat flour, with the best overall accuracy of 97.53%, regardless of the adulterated cassava flour variety. The best overall prediction accuracy of 93.83% was obtained in discriminating the flour samples into the three classes of 0% (pure wheat), 5% + 10% and 20% + 30% + 40%, regardless of the adulterated cassava flour variety. However, the highest overall accuracy of the 6-class model attained only 75.31% in classifying the wheat samples into the six groups of 0% (pure wheat), 5, 10, 20, 30 and 40%, regardless of the adulterated cassava flour variety. Overall, the obtained results demonstrated the usefulness of the employed low-cost spectrometer in detecting the wheat flour adulteration in a rapid and non-destructive manner.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have