Grade Identification of Tieguanyin Tea Using Fluorescence Hyperspectra and Different Statistical Algorithms

Area of Science:

• Agricultural Science
• Spectroscopy
• Machine Learning

Background:

• Traditional tea grading relies on manual inspection, which is time-consuming and subjective.
• Developing rapid, objective, and nondestructive methods for tea quality assessment is crucial for the tea industry.

Purpose of the Study:

• To develop and validate a fluorescence hyperspectral imaging (FHSI) method for accurate and nondestructive identification of Tieguanyin tea grades.
• To optimize machine learning models for improved tea grade classification accuracy.

Main Methods:

• Collected 309 Tieguanyin tea samples of three different grades.
• Acquired fluorescence hyperspectral data using a hyperspectrometer (400-1000 nm).
• Selected characteristic wavelengths using Bootstrapping Soft Shrinkage (BOSS), Variable Iterative Space Shrinkage Approach (VISSA), and Model Adaptive Space Shrinkage (MASS) algorithms.
• Established Support Vector Machine (SVM) models using full spectra, characteristic spectra, and selected wavelengths.
• Optimized SVM parameters using the Artificial Bee Colony (ABC) algorithm to create the VISSA-ABC-SVM model.

Main Results:

• The VISSA-SVM model showed promising classification performance.
• Optimization using the Artificial Bee Colony (ABC) algorithm significantly improved the VISSA-SVM model.
• The final VISSA-ABC-SVM model achieved a test set accuracy of 97.436% and a Kappa coefficient of 0.962.
• FHSI combined with the optimized SVM model demonstrated high accuracy in identifying Tieguanyin tea grades.

Conclusions:

• Fluorescence hyperspectral imaging (FHSI) coupled with the optimized VISSA-ABC-SVM model provides an accurate and rapid method for nondestructive Tieguanyin tea grade identification.
• This approach can be integrated into online tea grade detection systems, addressing limitations of existing methods.
• The developed method has practical applications for tea companies, markets, and farmers, enhancing tea quality assessment.