基于色差系统的宝庆桂丁红茶品质量化评价模型构建

目的:基于茶叶色差构建宝庆桂丁红茶品质量化评价模型。方法:以宝庆桂丁红茶为研究对象，在感官审评的基础上，分别测定干茶、茶汤和叶底的色差值，分析茶叶色差值与品质之间的相关性，并以GA-BP神经网络构建品质量化评价模型。结果:宝庆桂丁红茶品质与茶汤和叶底的L、a、b值呈极显著相关（P<0.01），与干茶a值呈显著相关（P<0.05）；遗传算法（GA）的引进明显提高了BP神经网络的拟合精度，GA-BP模型的决定系数（R2）明显高于BP网络；通过对比不同隐含层结构，优选结构为9-5-1的GA-BP神经网络结构；优选的GA-BP神经网络模型的训练、验证、测试和预测的决定系数（R2）分别为0.988、0.976、0.933和0.95。结论:基于色差系统和GA-BP神经网络对宝庆桂丁红茶品质进行量化评价的方法是可行的。

Construction of Quantitative Evaluation Model of Baoqing Guiding Black Tea Based on Chromatic Aberration Analysis

Objective: In order to construct a quantitative evaluation model of Baoqing Guiding black tea based on chromatic aberration. Methods: Baoqing Guiding black tea were used as materials. On the basis of sensory evaluation, the chromatic aberration value of dry tea, tea infusion and infused leaf were measured respectively. The correlation between chromatic aberration value and tea quality were analyzed, and the quality evaluation model were constructed by GA-BP neural networks. Results: The quality of Baoqing Guiding black tea was extremely significant correlated with the Lab value of tea infusion and leaves (P<0.01), and significantly correlated with the a value of dry tea (P<0.05). The introduction of genetic algorithm (GA) obviously improved the fitting accuracy of BP neural network, the coefficient of determination (R2) of GA-BP model was obviously higher than BP model. By comparing different hidden layer structures, the GA-BP neural network with structure of 9-5-1 were selected. The determination coefficients (R2) of training, verification, test and prediction of optimized GA-BP model were 0.988, 0.976, 0.933 and 0.95 respectively. Conclusion: The quantitative quality evaluation method of Baoqing Guiding black tea based on chromatic aberration and GA-BP neural network was feasible.