Scholarlink

<p>In the context of Tieguanyin oolong tea processing and sales, effectively identifying the degree of roasting and rapidly predicting its quality are crucial. However, traditional detection approaches are often time-consuming and require significant manual effort. In this study, data fusion of near-infrared spectroscopy (NIRS) and electronic nose (E-nose) technologies was employed to classify roasting levels and predict the quality of Tieguanyin oolong tea. NIRS is sensitive to chemical composition changes, while E-nose is responsive to variations in volatile compounds. Their combined application enhances the effectiveness of roasting degree discrimination. By comparing discriminant models based on single technologies and fusion strategies, it was found that the highest test set accuracy achieved under data fusion was 98.83%. Additionally, by integrating the intrinsic quality components, quality prediction models were developed by implementing data-level and feature-level fusion of information obtained from NIRS and E-nose data. The results showed that for free amino acids, tea polyphenols, and catechins, the low-level fusion method on the DZB-SNV dataset exhibited the best prediction performance, while for caffeine, the low-level fusion method on the DZB-MSC dataset performed optimally. These findings indicate that the data fusion strategy based on NIR and E-nose effectively compensates for missing information in the E-nose dataset, providing a novel and effective approach for roasting degree discrimination and quality prediction of Tieguanyin oolong tea.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Discrimination of the degree of roasting and quality prediction of Tieguanyin oolong tea based on NIRS, E-nose, and data fusion

  • 期刊论文

  • 发表日期：

    2026年1月30日

查看全文
  • Haokun Du,
  • Wencong Liu,
  • Lizhu Huang,
  • Feihu Song,
  • Zhenfeng Li,
  • Wanxiu Xu,
  • Chunfang Song

摘要

In the context of Tieguanyin oolong tea processing and sales, effectively identifying the degree of roasting and rapidly predicting its quality are crucial. However, traditional detection approaches are often time-consuming and require significant manual effort. In this study, data fusion of near-infrared spectroscopy (NIRS) and electronic nose (E-nose) technologies was employed to classify roasting levels and predict the quality of Tieguanyin oolong tea. NIRS is sensitive to chemical composition changes, while E-nose is responsive to variations in volatile compounds. Their combined application enhances the effectiveness of roasting degree discrimination. By comparing discriminant models based on single technologies and fusion strategies, it was found that the highest test set accuracy achieved under data fusion was 98.83%. Additionally, by integrating the intrinsic quality components, quality prediction models were developed by implementing data-level and feature-level fusion of information obtained from NIRS and E-nose data. The results showed that for free amino acids, tea polyphenols, and catechins, the low-level fusion method on the DZB-SNV dataset exhibited the best prediction performance, while for caffeine, the low-level fusion method on the DZB-MSC dataset performed optimally. These findings indicate that the data fusion strategy based on NIR and E-nose effectively compensates for missing information in the E-nose dataset, providing a novel and effective approach for roasting degree discrimination and quality prediction of Tieguanyin oolong tea.