Combination of diffuse optical reflectance spectroscopy and laser-induced breakdown spectroscopy for accurate classification of edible salts

Abstract

Diffuse optical reflectance spectroscopy (DORS) and laser-induced breakdown spectroscopy (LIBS) were applied for classification of edible salts. DORS and LIBS spectra were recorded for 21 edible salts products. Principal component analysis was performed to reduce variables for each spectroscopic data. DORS and LIBS classification models using principal component scores of the calibration dataset were constructed. Each model was validated for external test dataset using the k-nearest neighbors algorithm. The LIBS model provided better classification accuracy than the DORS model. However, the DORS model showed higher accuracy for particular salt samples. This indicates that reflectance spectra possess independent chemical information from that provided by LIBS spectra. Complementary effects of DORS and LIBS on improving the classification accuracy were investigated for the classification problem composed of selected samples. DORS-LIBS fused models were constructed with different weighting factors for DORS variables. The optimum fused model showed higher classification accuracy than those of either DORS or LIBS models. Our results suggest that the two simple spectroscopic techniques can be combined for highly accurate classification of edible salt products.