Classification of quality grading of Anji white tea using hyperspectral imaging and data fusion techniques

Anji white tea, a premium green tea renowned for its distinctive characteristics and considerable economic value, traditionally relies on sensory evaluation and chemical analysis for quality assessment. Sensory methods, though intuitive, are subjective and time consuming, while chemical techniques such as gas chromatography and high performance liquid chromatography, provide only a limited view of a few chemical components. This study introduces hyperspectral imaging (HSI) technology to evaluate Anji white tea quality by simultaneously capturing spectral and image data. Spectral data were preprocessed and analyzed to extract features related to catechins, tea polyphenols, and free amino acids. Image data texture features were extracted using the gray-level co-occurrence matrix, and significant features related to tea grades were identified through correlation analysis. Classification models using support vector machine and K-nearest neighbors (KNN) algorithms were developed. To enhance model accuracy, two data fusion strategies—low-level and mid-level—were applied. The results showed that both fusion strategies outperformed single-data models, with the KNN model using mid-level fusion achieving 100 % accuracy in both the training and test sets. This study demonstrates the potential of HSI and data fusion in improving the accuracy of tea quality assessment, providing a solid framework for the scientific evaluation of Anji white tea.