聚类算法用于中药材的近红外光谱分析

将近红外光谱分析技术结合聚类算法应用于中药材的鉴别分析,为当前中药材的类别分析提供了一种新的思路.选择黄樟油、桂叶油、桉叶油、松油醇、丁香叶油及三个不同来源的艾叶油共8个中药样品用实验室自行研制的多通道近红外光谱仪器进行快速(数秒钟)吸收光谱测试.以空气为本底,测试光谱范围700～1 700 nm的药品透射光谱(吸光度),结果显示8个样品的光谱差异较大.对此八种近红外光谱建立了定性数学模型,然后进行了聚类分析,并选择不同的聚类方法进行优化,得到相似系数为0.974 2,表明该聚类分析可行.在聚类分析产生的聚类树中,8个样品的聚类性质与其实际特性相符良好,特别是三种艾叶油被归为最近的类中,因此该聚类分析结果合理.

Clustering Analysis Applied to Near-Infrared Spectroscopy Analysis of Chinese Traditional Medicine

The present article discusses the clustering analysis used in the near-infrared (NIR) spectroscopy analysis of Chinese traditional medicines, which provides a new method for the classification of Chinese traditional medicines. Samples selected purposely in the authors' research to measure their absorption spectra in seconds by a multi-channel NIR spectrometer developed in the authors' lab were safrole, eucalypt oil, laurel oil, turpentine, clove oil and three samples of costmary oil from different suppliers. The spectra in the range of 0.70-1.7 microm were measured with air as background and the results indicated that they are quite distinct. Qualitative mathematical model was set up and cluster analysis based on the spectra was carried out through different clustering methods for optimization, and came out the cluster correlation coefficient of 0.9742 in the authors' research. This indicated that cluster analysis of the group of samples is practicable. Also it is reasonable to get the result that the calculated classification of 8 samples was quite accorded with their characteristics, especially the three samples of costmary oil were in the closest classification of the clustering analysis.