祁门红茶加工中氨基酸和儿茶素快速检测模型建立

氨基酸与儿茶素是茶叶品质的重要组成成分。祁门红茶在加工过程中，氨基酸与儿茶素含量发生了显著的变化，而且不同加工阶段差异性很大，但目前在生产中缺乏快速在线检测方法。为了实现对祁门红茶加工过程中氨基酸和儿茶素含量快速测定，试验以鲜叶、萎凋叶、揉捻叶、发酵叶和干燥后毛茶为原料，获取近红外光谱并利用化学方法检测氨基酸和儿茶素含量。对采集的原始光谱进行标准正态变量变换(SNVT)预处理，利用联合区间偏最小二乘回归法(Si-PLS)构建氨基酸和儿茶素含量近红外回归模型，相关系数与交互验证均方根误差作为评价模型的有效指标。结果表明，利用Si-PLS方法建立氨基酸含量的模型最优组合包含20个光谱区间并联合4个子区间和9个主成分因子，校正集的相关系数、校正均方根误差分别为0.955 8和1.768;预测集的相关系数、预测均方根误差分别为0.949 5和2.16。儿茶素含量的模型最优组合包含20个光谱区间并联合3个子区间和10个主成分因子，校正集的相关系数、校正均方根误差分别为0.940 1和1.22;预测集的相关系数、预测均方根误差分别为0.938 5和1.17。所建立模型准确性较好，这为茶叶加工过程中茶叶品质的在线监控提供了理论依据。

Rapid and Dynamic Determination Models of Amino Acids and Catechins Concentrations during the Processing Procedures of Keemun Black Tea

Tea is one of the most popular beverages in the world. For the contribution to the taste and healthy functions of tea, amino acids and catechins are important components. Among different kinds of black teas in the world, Keemun black tea has the famous and specific fragrance, “Keemun aroma”. During the processing procedure of Keemun black tea, the contents of amino acids and catechins changed greatly, and the differences of these concentrations during processing varied significantly. However, a rapid and dynamic determination method during the processing procedure was not existed up to now. In order to find out a rapid determination method for the contents of amino acids and catechins during the processing procedure of Keemun black tea, the materials of fresh leaves, withered leaves, twisted leaves, fermented leaves, and crude tea (after drying) were selected to acquire their corresponding near infrared spectroscopy and obtain their contents of amino acids and catechins by chemical analysis method. The original spectra data were preprocessed by the Standard Normal Variate Transformation (SNVT) method. And the model of Near Infrared (NIR) spectroscopy with the contents of amino acids and catechins combined with Synergy Interval Partial Least squares (Si-PLS) was established in this study. The correlation coefficients and the cross validation root mean square error are treated as the efficient indexes for evaluating models. The results showed that the optimal prediction model of amino acids by Si-PLS contained 20 spectral intervals combined with 4 subintervals and 9 principal component factors. The correlation coefficient and the root mean square error of the calibration set were 0.955 8 and 1.768, respectively; the correlation coefficient and the root mean square error of the prediction set were 0.949 5 and 2.16, respectively. And the optimal prediction model of catechins by Si-PLS contained 20 spectral intervals combined with 3 subintervals and 10 principal component factors. The correlation coefficient and the root mean square error of the calibration set were 0.940 1 and 1.22, respectively; the correlation coefficient and the root mean square error of the prediction set were 0.938 5 and 1.17, respectively. The results showed that the established models had good accuracy which could provide a theoretical foundation for the online determination of tea chemical components during processing.