Geographical Origin Identification of Dendrobium officinale Using Variational Inference-Enhanced Deep Learning.

Area of Science:

• Agricultural Science
• Computer Science
• Pharmacology

Background:

• Dendrobium officinale is a valuable medicinal plant in China.
• Origin significantly impacts its quality, necessitating accurate identification.
• Current methods are subjective or expensive, lacking efficiency.

Purpose of the Study:

• To develop a high-precision, non-destructive method for Dendrobium officinale origin identification.
• To address the challenge of accurate geographical origin identification with limited image samples.
• To advance intelligent quality assessment for Dendrobium officinale.

Main Methods:

• Proposed a Variational Inference-enabled Data-Efficient Learning (VIDE) model.
• VIDE utilizes dual probabilistic networks for feature extraction and classification.
• Employed variational inference to model feature distributions for robust classification.

Main Results:

• The VIDE model achieved 91.51% precision, 92.63% recall, and 92.07% F1-score.
• Demonstrated superior performance compared to state-of-the-art models on a custom dataset.
• Successfully identified samples from six major Chinese regions with high accuracy.

Conclusions:

• The VIDE model offers a practical and efficient solution for Dendrobium officinale origin identification.
• This approach enables accurate, non-destructive quality assessment using limited image data.
• Advances intelligent systems for traditional Chinese medicine authentication.