Abstract
Tea is the second most popular beverage globally after water. Identifying fresh tea leaves is significant to ensure product quality in tea industry, since they are often difficult to distinguish via natural colors and morphological features. Traditional component analysis methods of fresh tea leaves are destructive and costly, while emerging non-destructive techniques typically rely on large-scale and time-consuming instruments, limiting their applicability in on-site scenarios. In this study, we propose a portable identification system for fresh tea leaves, which is the combination of a field-spectrometer and an advanced deep learning (DL) architecture for spectral classification. A custom-designed leaf fixation apparatus is introduced to ensure stable and reliable spectra acquisition, enabling robust data acquisition for DL model training in field environments. The DL model adopts a Transformer architecture enhanced by principal component analysis, which not only reduces 99.4 % of training parameters compared to conventional DL methods, but also elevates tea leaf classification accuracy. As a proof of concept, we apply the proposed detection system to identify the Wuyi Rock Tea, a world-renowned type of Chinese tea with the unique flavour profile and rich cultural heritage. Our approach achieves classification accuracies of 99.15 % for tea variety and 100 % for tea quality, outperforming several existing methods. This study provides a convenient solution for rapid identification of fresh tea leaves, and will also highlight the broader potential of our scheme on other leaf-identification-based applications.
[Display omitted]
•We develop a field-portable spectrometer, identifying a fresh tea leaf non-destructivelyrapidly.•The artificial intelligence model reduces training parameters by 99.4 % and increases accuracy by 90.2 % than the conventional method.•The system accomplishes the variety and quality accuracies up to 99.2 % and 100.00 %, respectively.•The system exhibits a great potential on convenient quality control of fresh leaves in tea industry.