基于电子舌技术和模式识别方法的茶叶质量等级评判

Discrimination of tea’s quality level based on electronic tongue and pattern recognition

目的:尝试利用电子舌技术来评判茶叶等级,以提高评判结果的客观性和公正性;方法:试验以4个等级的炒青绿茶为研究对象,对获取的电子舌数据,利用K最近邻域(KNN)模式识别方法建立茶叶等级质量的评判模型,在模型建立过程中,模型参数K和主成分因子数(PCs)通过交互验证的方法被优化;结果:在K=1和PCs=5时,所得到的模型最佳,模型交互验证识别率为97.5%,对预测集中样本进行验证时,预测识别率为100%;结论:电子舌技术与适当的模式识别方法相结合可以成功地评判茶叶的质量等级。

Objective： To improve the objective results in estimating tea quality, electronic tongue as a new sensors analytical tool was used to evaluate tea quality coupled with an appropriate pattern recognitlon method in this study ; Methods： Four grades of roasted green tea were studied in the experiment. K-Nearest Neighbors （KNN）were applied to build discriminating model as a pattern recognition method, parameter K of the KNN model and the number of principal component factors （PCs） were optimized by cross-validation in building models; Results： Experimental results showed that the optimal model was obtained with PCs = 5 and K = 1, and the discriminating rates equal to 97.5% and 100% in training and validation set, respectively ; Conclusion ： The overall results demonstrated that electronic tongue technology with an appropriate pattern recognition method can be successfully applied to evaluate tea quality level.