Research on the adulteration of Lycium barbarum based on hyperspectral imaging technology combined with deep learning algorithms

Abstract

Consumers favour the exceptional medicinal and edible qualities of goji berry from Zhongning in Ningxia. Nevertheless, the market frequently witnesses the blending of goji berry from other regions in Ningxia, necessitating the urgent need for an effective adulteration detection method. The study employs hyperspectral imaging technology coupled with chemometrics to accurately detect adulteration in goji berry. Hyperspectral images of goji berry samples were collected from five regions in Ningxia: Zhongning, Huinong, Tongxin, Haiyuan, and Helan. Zhongning goji berry were designated as the adulteration targets for goji berry from the region of Huinong, Tongxin, Haiyuan, and Helan, with adulteration rates of 10%, 20%, and 30%. Through the division of the sample sets and correlation analysis, 20% adulteration rate was ultimately chosen for spectral analysis. The study revealed that the RF model demonstrated superior accuracy compared to other methods, achieving accuracy of 95.8%, 91.4%, 87.2%, and 91.1%, respectively. Hence, the RF model is considered the preferred choice for accurately identifying the adulteration of 20% wolfberry from diverse regions in Ningxia.