基于FT-IR结合PCA-LDA的白酒真伪鉴别方法
Identification of liquor authenticity based on FTIR with PCA- LDA
-
摘要: 运用傅里叶变换红外光谱法（FT-IR）采集白酒的红外光谱图,并采用主成分分析法（Principal Component Analysis,PCA）对白酒红外光谱数据进行降维处理,提取累积方差贡献率达到99%以上的主成分作为建模参数,结合线性判别分析法（Linear Discriminant Analysis,LDA）建立基于白酒红外光谱全波段、指纹区和一阶导数谱的白酒真伪鉴别模型。结果显示,采用白酒红外光谱全波段数据所建立的白酒真伪鉴别模型的精度更高,其初始判别准确率为95.6%,交叉验证实验中的判别准确率为91.4%,对测试集中30组样本的判别正确率为86.7%。本研究可以为白酒的真伪鉴别提供技术参考。Abstract: The infrared spectrograms of liquor samples were gathered by FT- IR and principal component analysis was used for datas dimension- reduced processing. The principal components which cumulative variance contribution rate exceed 99% were extracted as the modeling parameters,combined with the discriminant analysis,the liquor authenticity identification model was established based on all wavelengths,fingerprint region and first derivative spectrum of liquors infrared spectroscopy,respectively.The results showed that the discriminant rate of identification model which was established by the all wavelengths data was the highest,the initial discrimination rate was 95.6% and the cross validation test of discriminant rate was 91.4%. The correct discrimination rate of this identification model for the 38 samples in validation set was 86.7%. This study could provide theoretical reference for liquor identification.
下载: