ScholarWorks@경상국립대학교

초록

In South Korea, the cultivation of Dendranthema grandiflorum, a plant used in ornamental purposes, food additives, and medicine, has decreased owing to increased production costs. Thus, the import of cheaper D. grandiflorum flowers from China and Vietnam is increasing drastically to meet domestic needs. However, current methods for discerning the country of origin of these plants and preventing illegal and false sales are limited and complex. The present study aimed to develop a classification model using Fourier transform nearinfrared spectroscopy data combined with partial least squares discriminant analysis (PLS-DA) to distinguish between D. grandiflorum varieties of different origins. A total of 422 D. grandiflorum samples, harvested from 2021 to 2023, were collected. The use of OPUS 7.0 for automatic searches for optimal pre-treatment and spectral regions facilitated the rapid development of high-quality PLS-DA models. Both a single-year PLS-DA model and an updated version for identifying geographical origins were examined to assess the impact of data augmentation on predictive accuracy. External validation revealed that the updated model yielded a notably higher prediction rate of 98.1%, compared to the 62.8% rate of the previous model, primarily due to data enrichment. The model accurately differentiated among the Jinba, Baekgang, Iwano-hakusen, and Seikono-makoto varieties. The method developed in the present study allows for the precise and simple concurrent identification of both the origin and varieties of D. grandiflorum with a single sample analysis, ultimately aiding in the prevention of adulteration and protection of plant varieties.

키워드