当前位置: 首页 > 文章 > 基于傅里叶红外光谱的多茶类判别研究 茶叶科学 2014 (1) 67-74
Position: Home > Articles > Discriminant Analysis of Tea Category Based on Fourier Infrared Spectra Journal of Tea Science 2014 (1) 67-74

基于傅里叶红外光谱的多茶类判别研究

作  者:
吴全金;董青华;孙威江
单  位:
福州市华茗茶业研究所;福建农林大学安溪茶学院;福建农林大学园艺学院
关键词:
茶类;傅里叶红外光谱;逐步判别分析
摘  要:
采用乌龙茶、绿茶、红茶和白茶成品茶样本,按同步建模和特定茶类建模两种策略,进行了基于傅里叶红外光谱和逐步判别分析技术的茶类判别研究。结果表明,以1 000~1 800 cm-1波段吸收值、主要吸收峰比值、萃取前二者的建模因子进行同步建模,所得回代判别准确率和外部验证准确率分别为84.8%、85.0%,81.9%、80.0%,89.0%、88.5%;3种数据处理方法针对特定茶类建模的回代判别准确率和外部验证准确率分别为94.2%、95.0%,89.2%、90.0%,93.0%、95.0%;同步建模和特定茶类建模分别得到判别效果较好的函数4条,能有效区分4种茶类,并且特定茶类建模的效果优于同步建模。
译  名:
Discriminant Analysis of Tea Category Based on Fourier Infrared Spectra
作  者:
WU Quanjin;DONG Qinghua;SUN Weijiang;College of Horticulture, Fujian Agriculture and Forestry University;. Fuzhou Huaming Tea Research Institute;Anxi Tea College, Fujian Agriculture and Forestry University;
关键词:
tea category,fourier infrared spectra,stepwise discriminant analysis
摘  要:
The discrimination of tea category based on FTIR and stepwise discriminant analysis was studied, using the synchronous modeling and specific tea modeling with Oolong tea, green tea, black tea and white tea as the samples. The results showed that the recognition accuracy of the resubstitution and external validation of tea category using synchronous modeling based on absorption values at 1 000~1 800 cm-1 band, the absorption ratio in the main peak, the extracted modeling factors of the former two were 84.8% and 85.0%, 81.9% and 80.0%, 89.0% and 88.5% respectively. The recognition accuracy of the resubstitution and external validation of tea category using specific tea modeling based on the three methods were 94.2% and 95.0%, 89.2% and 90.0%, 93.0% and 95.0% respectively, and 4 discriminant functions of synchronous analysis and specific tea analysis which can discriminate these four kinds of tea effectively were established respectively. The results also indicated that the effect of tea-specific modeling was better than synchronous modeling.

相似文章

计量
文章访问数: 11
HTML全文浏览量: 0
PDF下载量: 0

所属期刊

推荐期刊