欢迎访问《茶叶科学》,今天是

茶叶科学 ›› 2017, Vol. 37 ›› Issue (3): 258-265.

• • 上一篇    下一篇

基于嗅觉可视化技术的工夫红茶发酵程度判定方法

陈琳1,2, 叶阳1,*, 董春旺1,*, 何华锋1   

  1. 1. 中国农业科学院茶叶研究所,浙江省茶叶加工工程重点实验室,国家茶产业工程技术研究中心,农业部茶树生物学与资源利用重点实验室,浙江 杭州 310008;
    2. 中国农业科学院研究生院,北京 100081
  • 收稿日期:2016-08-16 修回日期:2016-10-13 出版日期:2017-06-15 发布日期:2019-08-22
  • 通讯作者: *yeyang@tricaas.com,dongchunwang@tricaas.com
  • 作者简介:陈琳,男,硕士研究生,主要从事茶叶加工方面的研究。
  • 基金资助:
    浙江省自然科学基金(LY16C160002)、浙江省重点研发计划(2015C02001)、国家星火计划(2015GA700006)

Monitoring Black Tea Fermentation Using a Colorimetric Sensor Array-based Artificial Olfaction System

CHEN Lin1,2, YE Yang1,*, DONG Chunwang1,*, HE Huafeng1   

  1. 1. Tea Research Institute, Chinese Academy of Agricultural Sciences, Key Laboratory of Tea Processing Engineering of Zhejiang Province, National Engineering Technology Research Center of Tea Industry, Key Laboratory of Tea Biology and Resource Utilization of Ministry of Agriculture, Hzngzhou 310008, China;
    2. Graduate School of Chinese Academy of Agricultural Sciences, Beijing 100081, China
  • Received:2016-08-16 Revised:2016-10-13 Online:2017-06-15 Published:2019-08-22

摘要: 发酵是工夫红茶加工的关键工序,对红茶品质形成起着极其重要的作用。本文提出一种基于嗅觉可视化技术的工夫红茶发酵程度判定方法。基于硅胶薄层层析板与16种卟啉衍生物设计构建了嗅觉可视化传感器及气体检测系统,用于工夫红茶发酵过程中挥发性气体数据采集。采用Fisher判别分析与BP-AdaBoost算法建立工夫红茶发酵程度判别模型。分析表明,Fisher判别函数可以实现不同发酵程度红茶100%分类,交叉验证分组正确率达90.74%;BP-AdaBoost算法建立判别模型,训练集相关系数(Rc)和预测集相关系数(Rp)分别为0.9578和0.9132;嗅觉可视化技术可以实现工夫红茶发酵程度判定,为工夫红茶发酵过程实时监控提供了理论依据。

关键词: 红茶, 发酵程度, Fisher判别, BP-AdaBoost算法

Abstract: As the crucial procedure for production of black tea, fermentation plays an important role in quality control of black tea. This paper proposed a colorimetric sensor array-based artificial olfaction system to monitor black tea fermentation. Herein, a colorimetric sensor array by printing 16 chemical dyes including porphyrins/metalloporphyrins on a Silica gel thin-layer chromatography plate was utilized to detect volatile gases during black tea fermentation. Discrimination model was established by fisher discriminatory analysis and adaptive boosting algorithm based on BP-ANN (BP-Adaboost). Results showed that the discrimination rate and discrimination rate of cross-validation reached 100% and 90.74% respectively. BP-Adaboost model showed that the correlation coefficient of calibration set (Rc) and prediction set (Rp) were 0.9578 and 0.9132 respectively. This work demonstrates that it is feasible to distinguish the degree of black tea fermentation using a colorimetric sensor array-based artificial olfaction system.

Key words: black tea, fermentation degree, Fisher discriminatory analysis, BP-Adaboost algorithm

中图分类号: