Improvement on the discrimination of Amomum tsao-ko origins using NIR combined with sequential pre-processing through orthogonalization (SPORT) and PLS-DA

Abstract

Amomum tsao-ko Crevost et Lemaire (A. tsao-ko) is mainly used as spices in Chinese hotpot and meat cooking for improving taste and removing peculiar smell. To guarantee the quality of A. tsao-ko, the demand of quick and efficient analysis method is in urge. In this study, A. tsao-ko from 5 regions (including 248 individuals) were analyzed by NIR. Furthermore, sequential preprocessing through orthogonalization (SPORT) was performed to fuse the complementary information of several scatter correction techniques for obtaining the high-quality model. The result showed that in the best results obtained on the scatter correction modeling, the accuracy values of training set and test set were respectively identical to 0.8 and 0.84. However, SPORT method can correctly discriminate all the samples. Hence, SPORT model outperformed the single scatter correction modeling. Furthermore, compared with the single scatter correction modeling, SPORT could not only take the user out of the loop of the selection of the best pre-processing but also could give insight into the incremental contributions of each processed NIR data. Therefore, SPORT could be used as a reliable method for quality control of A. tsao-ko.