The Fermentation Degree Prediction Model for Tieguanyin Oolong Tea Based on Visual and Sensing Technologies

Area of Science:

• Food Science and Technology
• Agricultural Engineering
• Machine Learning Applications

Background:

• Oolong tea fermentation quality control relies on subjective sensory evaluation.
• Objective, real-time monitoring methods are needed for efficient tea production.

Purpose of the Study:

• To develop and optimize machine learning models for assessing Tieguanyin oolong tea fermentation degree.
• To evaluate the effectiveness of single-feature, multi-source feature fusion, and algorithm optimization.

Main Methods:

• Collected data on water loss, aroma, color, and texture using sensors.
• Applied Support Vector Regression (SVR), Random Forest (RF), and Long Short-Term Memory (LSTM) models.
• Utilized data fusion techniques and the Sparrow Search Algorithm (SSA) for optimization.

Main Results:

• Single-feature models achieved R² values up to 0.959.
• Data fusion models significantly improved performance (R² up to 0.991).
• SSA-optimized models, particularly Fusion-SSA-LSTM, showed the best accuracy (R² up to 0.994).

Conclusions:

• Multi-source feature fusion enhances oolong tea fermentation assessment.
• Sparrow Search Algorithm optimization further improves model precision.
• Developed models enable online, real-time monitoring for automated Tieguanyin oolong tea production.