近红外光谱法结合化学计量学方法用于茶油真伪鉴别分析

摘要

采用近红外光谱透反射模式结合化学计量学方法对纯茶油进行真伪鉴别.收集并扫描了163个样品(合格97个,不合格66个),对样本进行光谱数据预处理优化及有效波段筛选.在5750～6000 cm-1波段处,光谱经过平滑,一阶导数以及自归一化后,采用无监督学习算法即主成分分析法(Principal component analysis,PCA)进行分类,然后再采用有监督学习算法即判别分析(Discriminant analysis,DA)建立校正模型,进行预测.PCA和DA都能够得到满意的结果,两种方法的分类准确率均达到98.8%.结果表明:近红外光谱可作为一种简单、快速、无损、可靠的方法用于鉴别纯茶油的真伪.%Near infrared spectroscopy(NIRS) in the transreflection mode combined with chemometrics was used to identify Camellia oil. The samples set contained 163 spectra of qualified (n＝97) and unqualified (n=66) have been collected and scanned, the spectral data were pretreated and selected for effective variables. In the wavenumber range 5750 cm-1 to 6000 cm-1 , the optimal combination of pretreatments (smoothed, first derivative, and autoscaling) was adopted to process the spectra.Unsupervised classification-principal component analysis (PCA) was firstly adopted to classify, and then supervised classification-discriminant analysis (DA) was used to build calibration model to predict. Satisfactory results were obtained by PCA and DA, the correct recognition rates of two methods can reach 98.8％. The results showed that NIRS can be used as a simple, rapid, nondestructive and reliable method to identify Camellia oil.