Small-Sample Authenticity Identification and Variety Classification of Anoectochilus roxburghii (Wall.) Lindl. Using...

Area of Science:

• Agricultural Science
• Computer Science
• Botany

Background:

• Accurate identification of medicinal plants like Goldthread (Anoectochilus roxburghii) is crucial.
• Counterfeit species pose a significant challenge in herbal medicine and trade.
• Hyperspectral imaging offers detailed spectral information for plant analysis.

Purpose of the Study:

• To develop and evaluate machine learning models for authenticating Anoectochilus roxburghii.
• To differentiate Goldthread from its counterfeit species using hyperspectral data.
• To explore the efficacy of various machine learning algorithms and spectral fusion techniques.

Main Methods:

• Collected hyperspectral data from the front and back leaves of nine Anoectochilus roxburghii species and two counterfeit species.
• Applied machine learning models: Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Random Forest (RF), Linear Discriminant Analysis (LDA), and Convolutional Neural Networks (CNN).
• Developed a multi-view spectral fusion Convolutional Neural Network (CNN) model integrating data from both leaf sides.

Main Results:

• Support Vector Machine (SVM) achieved 100% classification accuracy in distinguishing Goldthread from counterfeits.
• SVM demonstrated superiority in handling high-dimensional spectral data compared to traditional models.
• The multi-view spectral fusion CNN model also achieved a perfect 100% classification accuracy.
• Spectral differences between front and back leaves were effectively captured by the models.

Conclusions:

• Hyperspectral imaging combined with machine learning provides a highly effective method for plant authenticity identification.
• The developed SVM and multi-view spectral fusion CNN models offer robust solutions for detecting counterfeit species.
• This approach presents a novel and promising perspective for quality control in herbal products.