1. Counterfeit detection of bulk Baijiu based on fluorescence hyperspectral technology and machine learning.
- Author
-
Wu, Youli, Li, Xiaoli, Xu, Lijia, Fan, Rongsheng, Lin, Yi, Zhan, Chunyi, and Kang, Zhiliang
- Subjects
MACHINE learning ,ADULTERATIONS ,FLUORESCENCE ,PRINCIPAL components analysis ,PRODUCT counterfeiting ,FORGERY ,MULTISPECTRAL imaging ,SPECTRAL imaging - Abstract
Baijiu is a unique distilled spirit in China. The bulk Baijiu market has been experiencing issues related to counterfeit and substandard products, raising concerns about food safety. Detecting liquor adulteration is crucial for eliminating fraud in the bulk Baijiu market. In this study, we proposed using fluorescence hyperspectral Technology (FH) combined with machine learning (ML) to detect Baijiu adulteration quickly and non-destructive. Due to the similarity of fluorescence spectral features between adulterated Baijiu and real Baijiu, it was difficult to distinguish them based on the fluorescence feature parameters alone. The data preprocessing methods were used and then principal component analysis (PCA) was adapted. The principal components were used as inputs to ML models to establish the qualitative and quantitative detection models. In the qualitative detection models, the Adaptive Boosting (AdaBoost) model demonstrated the best performance with 98.08% precision, 100% recall and 99.03% F1-score. In the quantitative detection models of adulterations concentration, the AdaBoost model after Wavelet denoising(WDS) processing yielded the best results with R
2 of 0.9740 and RMSEP of 0.0247. The results demonstrated that the combination of FH and ML can efficiently detect adulterated bulk Baijiu, showing promising applications and feasibility in the nondestructive detection of adulterated substances. [ABSTRACT FROM AUTHOR]- Published
- 2024
- Full Text
- View/download PDF