Non-destructive discrimination of ginseng, American ginseng, and Korean ginseng based on hyperspectral imaging

Ginseng, American ginseng, and Korean ginseng share similar appearances but exhibit significant differences in their efficacy, composition, and market value. Traditional identification methods for distinguishing between them are highly subjective and inefficient. To achieve rapid, accurate, and non-destructive species identification, this study employed 900–1700 nm hyperspectral imaging technology combined with chemometric modeling. This study is the first to systematically compare the spectral characteristics of the three ginseng types in the near-infrared band, and proposes an optimized modeling strategy integrating Savitzky–Golay combined with first-order derivative (SG1) preprocessing and feature wavelength selection. A total of 150 samples were collected and divided into a training set (105 samples) and a test set (45 samples) at a 7:3 ratio. Results demonstrated that SG1 combined with either CARS or UVE enabled multiple models (BPNN, ELM, PLS-DA) to achieve 100% classification accuracy on the test set. Furthermore, this study pioneered the generation of pseudo-color images through principal component analysis, enabling visualization of the three spectral features in image space and providing intuitive support for variety identification. This method offers an efficient, non-destructive technical pathway for variety identification and quality control of high-value Chinese medicinal materials, and demonstrates promising application prospects. • Ginseng types are traditionally difficult to differentiate. • Hyperspectral imaging combined with chemometric models was examined. • Hyperspectral images of three ginseng types were acquired. • Savitzky-Golay first derivative and feature selection enhanced model performance. • The method enables precise and non-destructive differentiation of ginseng types.