Non-destructive geographical origin authentication of licorice using hyperspectral imaging and a component-aware Swin...

Area of Science:

• Agricultural Science
• Food Science
• Computer Science

Background:

• Geographical origin authentication is vital for medicinal and functional foods like licorice (Glycyrrhiza uralensis Fisch.).
• Distinguishing origins is challenging due to subtle compositional variations and overlapping spectral signatures in licorice samples.
• Non-destructive authentication methods are needed for quality control and traceability.

Purpose of the Study:

• To develop a deep learning framework for non-destructive, accurate geographical origin authentication of licorice.
• To enhance the performance of hyperspectral imaging (HSI) analysis for complex food materials.
• To improve the reliability of food origin traceability and quality assurance.

Main Methods:

• A hyperspectral imaging (HSI)-based deep learning framework, Component-Aware Swin Transformer (CAST), was proposed.
• Sliding-window local principal component analysis (PCA) with the Landgrebe criterion reduced spectral redundancy.
• A PCA-aware multi-scale feature extractor with a Component-Aware Squeeze-and-Excitation mechanism was utilized.
• A 3D Swin Transformer block captured spectral-spatial dependencies, optimized with Center Loss and Label Smoothing.

Main Results:

• The CAST model achieved an overall accuracy of 95.22% on 1045 licorice samples from three regions.
• CAST outperformed the strongest baseline model, HyperSFormer, by 2.54 percentage points.
• The results indicate effective feature extraction and classification for origin authentication.

Conclusions:

• The proposed CAST framework demonstrates significant potential for reliable food origin authentication using HSI and deep learning.
• This approach offers a non-destructive and intelligent method for quality control of medicinal and functional food materials.
• Integrating advanced deep learning with HSI provides a powerful tool for ensuring the safety and traceability of agricultural products.