Rapid and non-destructive origin authentication of goji berries using Raman spectroscopy and a lightweight deep learning framework

Zhiqing Yang, Min Chen, Shumin Gao, Rongxuan Wu, Haofan Zhang, Mengya Zhang, Yu Yang, Yao Qin, Peng Li,
Rapid and non-destructive origin authentication of goji berries using Raman spectroscopy and a lightweight deep learning framework,
Journal of Food Composition and Analysis,
Volume 148, Part 4,
2025,
108651,
ISSN 0889-1575,
https://doi.org/10.1016/j.jfca.2025.108651.
(https://www.sciencedirect.com/science/article/pii/S088915752501467X)
Abstract: Goji berries (Lycium barbarum) are widely recognized for their therapeutic and nutritional benefits, yet their chemical composition and efficacy vary significantly by geographic origin. Ensuring authenticity is crucial for quality control and preventing adulteration. In this study, a rapid and non-destructive identification method was developed by integrating Raman spectroscopy with an optimized lightweight deep learning model, BGS-DenseNet. This framework enhances spectral feature extraction and fusion while maintaining low computational cost. Experiments using samples from multiple origins demonstrated that the proposed model achieved 99.65 % classification accuracy, outperforming conventional machine learning and standard convolutional neural networks. Due to its efficiency and ease of deployment, the method is highly applicable to real-time, field-based origin tracing of herbal and agricultural products. This study offers a scalable and intelligent solution for traceability in traditional Chinese medicine and food safety management.
Keywords: Raman spectroscopy; Origin authentication; Goji berry; Lightweight deep learning; BGS-DenseNet