Multimodal deep learning with hyperspectral imaging for accurate origin classification of wolfberries

Area of Science:

• Agricultural Science
• Computer Science
• Food Science

Background:

• Accurate classification of wolfberry geographical origin is crucial for its nutritional and medicinal value.
• Traditional methods for origin classification often lack efficiency and accuracy.

Purpose of the Study:

• To develop a highly accurate method for classifying wolfberry geographical origin.
• To compare the performance of a multimodal deep learning model against traditional machine learning approaches.

Main Methods:

• A multimodal convolutional neural network (MTCNN) with a cross-attention mechanism was developed.
• Spectral and image features were extracted and fused using the MTCNN.
• Traditional methods including Principal Component Analysis (PCA), Support Vector Machine (SVM), and K-Nearest Neighbor (KNN) were used for comparison.

Main Results:

• The MTCNN achieved a test accuracy of 99.88% for wolfberry origin classification.
• The best traditional model (SVM with GBDT-extracted spectral features and fused image data) reached 96.68% accuracy.
• The multimodal deep learning approach significantly outperformed conventional methods.

Conclusions:

• Multimodal deep learning offers superior performance for agricultural product traceability and authentication.
• The proposed MTCNN model is efficient, interpretable, and suitable for practical applications in quality assurance.
• This study demonstrates the potential of advanced AI techniques in enhancing food safety and product verification.