基于电子鼻技术的湄潭春茶产地溯源研究

Study on Origin Tracing of Meitan Spring Tea Based on Electronic Nose Technology

为实现不同产地湄潭春茶的快速、客观判别,基于电子鼻与多元统计分析定性判别不同产地湄潭春茶间的差异,并定量预测其产地。对电子鼻信号进行分析,发现其响应信号在传感器S7、S9、S6和S2的强度均较明显;方差分析发现湄潭春茶产地对传感器响应影响均显著;基于第80 s数据进行主成分分析(PCA)和典则判别分析(CDA),发现其基本能区分不同产地的湄潭春茶,且数据点的分布均与各产地地理位置分布呈现一定特征规律性,但区分效果不够理想。为进一步提高对不同乡镇茶叶的区分效果,利用不同特征值(平均值、曲线面积、最大值、斜率、主成分分析优选参数、loading优选参数)进行判别分析,其中平均值、曲线面积、最大值判别结果优于80 s时的判别结果;斜率、主成分分析优选参数、loading判别结果比80 s时的判别结果差;以响应曲线的平均值进行判别时,可达到100%正确识别。多层神经网络分析(MLP)作为效果最佳、决定系数最高(R_(c)^(2)=0.9855,R_(p)^(2)=0.9941)的茶叶产地预测模型,可以实现对不同产地湄潭春茶的有效预测。因此,以特征值中平均值数据作为输入,采用电子鼻结合多元统计与神经网络分析,可以实现对春茶产地的判别,以期为湄潭春茶开发利用、产地追溯、真伪鉴定提供科学依据。

In order to achieve rapid and objective discrimination of Meitan spring tea from different producing areas,the differences between Meitan spring tea from different producing areas are qualitatively discriminated based on electronic nose and multivariate statistical analysis,and the producing areas are quantitatively predicted.By analyzing the signal of electronic nose,it is found that the intensity of response signal in S7,S9,S6 and S2 sensors is obvious.Variance analysis shows that the producing area of Meitan spring tea has significant influence on the sensor response.Principle Component Analysis(PCA)and Canonical Discriminant Analysis(CDA)were carried out based on the data collected at the time of 80 s,and it is found that the method can basically distinguish Meitan Spring tea from different producing areas,and the distribution of data points shows certain characteristics and regularity with the geographical distribution of each producing area,but the discrimination effect is not ideal.In order to further improve the discrimination effect of tea leaves from different towns,different characteristic values(mean value,curve area,maximum value,slope,principal component analysis optimization parameter and loading optimization parameter)are used for discriminant analysis.The discriminant results of mean value,curve area and maximum value are better than those obtained at the time of 80 s.The discriminant results of slope,principal component analysis and loading are worse than those at the time of 80 s.When discriminating with the average value of the response curve,100%correct recognition can be achieved.Multilayer Perceptron(MLP)is the best prediction model with the highest determination coefficient(R_(c)^(2)=0.9855,R_(p)^(2)=0.9941),which can realize the effective prediction of Meitan Spring tea from different regions.Therefore,with the average data of characteristic values as input,electronic nose combined with multivariate statistics and neural network analysis can be used to identify the origin of spring tea,providing scientific basis for the development and utilization,origin tracing and authenticity identification of Meitan spring tea.