定标集样品数对茶叶近红外光谱分析精度的影响

Influence of calibration sets number on NIR model of tea

为了考察定标集样品数量对近红外光谱分析结果的影响,以2007年不同地区的高、中、低档茶叶为试验材料,采用MWPLS方法优化波长并在不同数量的定标集下分别建立茶多酚和咖啡碱组分的预测模型.结果表明:茶叶中茶多酚和咖啡碱组分模型的预测精度与建模样品的数量存在一定的关系;随着定标集样品数量的不断增加,模型的相关系数由小变大逐渐趋于平稳;在160个建模样品中,当茶多酚和咖啡碱定标集数量分别为85和100时,模型的相关系数最大,分别为0.974 1和0.963 7.

In order to study the influence of the number of calibration sets on the NIR veracity, MWPLS method was used to select wavelength, then the tea polyphenols and caffeine NIR models were established respectively for the different number of calibration sets by various tea of 2007. The results showed that the number of calibration sets had an obvious influence on tea polyphenols and caffeine models. With the number of calibration sets increasing, the predicted correlation coefficient （R） increased and finally became steady. When the numbers became 85 and 100 respectively, the NIR model veracity were highest which reached 0,974 1 and 0. 963 7.