论文部分内容阅读
本研究使用拉曼光谱分析技术采集不同产地和不同酒龄的黄酒样品指纹信息,对比判别分析(DA)和最小二乘支持向量机(LS-SVM)所建黄酒品质快速模型性能,确定最优模型以实现快速准确地评价黄酒品质。本研究在全波段范围利用主成分分析对拉曼光谱数据降维,计算降维谱图间马氏距离,基于ward’s算法建立判别分析模型;将全波段范围作为最小二乘支持向量机的输入量,选择出能较好处理非线性问题的RBF为核函数,同时采用交互验证方式优化RBF核函数参数,基于优化RBF核函数,建立最小二乘支持向量机鉴别模型。研究结果表明:拉曼光谱结合最小二乘支持向量机鉴别模型对黄酒产地和酒龄的鉴别正确率均为100%;拉曼光谱结合判别分析鉴别模型对嘉善、绍兴和上海黄酒的鉴别正确率分别为100%、80%和80%,对黄酒酒龄的鉴别正确率均为100%;最小二乘支持向量机模型性能优于判别分析模型。拉曼光谱结合化学计量学方法可快速、准确评价黄酒品质。
In this study, Raman spectroscopy was used to collect fingerprints of rice wine samples from different regions of origin and different wine ages. By comparing the performance of fast wine model with discriminant analysis (DA) and least squares support vector machine (LS-SVM) Model to achieve fast and accurate evaluation of rice wine quality. In this study, principal component analysis (PCA) was used to reduce the Raman spectrum data in the whole band range, and the Mahalanobis distance between the reduced dimension spectra was calculated. The discriminant analysis model was established based on ward’s algorithm. The full band range was taken as the input of least square support vector machine , Select RBF which can deal with nonlinear problem well as kernel function, optimize the parameters of RBF kernel function by interactive verification method, and establish least square support vector machine discriminant model based on optimized RBF kernel function. The results show that: Raman spectroscopy combined with least square support vector machine identification model for wine production and wine age were 100% correct identification rate; Raman spectroscopy combined with discriminant analysis model of Jiashan, Shaoxing and Shanghai rice wine identification accuracy Respectively, 100%, 80% and 80% respectively, and the correct identification rate of the wine age was 100%. The least squares support vector machine model was better than the discriminant analysis model. Raman spectroscopy combined with chemometric methods can quickly and accurately evaluate the quality of rice wine.