论文部分内容阅读
目的:提出了一种用紫外-可见光谱技术快速鉴别4种唇形科药材(薄荷、荆芥、香薷及广藿香)的新方法。方法:首先用主成分分析法对4种药材进行聚类分析并获取他们的紫外-可见指纹图谱,再结合人工神经网络技术进行品种鉴别。结果:主成分分析表明,主成分1和主成分2的累积可信度已达99%,以主成分1和2对所有建模样本的得分值做出的得分图,对4种药材具有良好的区分作用。利用对于4种药材品种敏感的特征波段作为神经网络的输入建立三层BP人工神经网络模型。对未知的20个样本进行预测,品种识别准确率达到100%。结论:该方法具有很好的分类和鉴别作用,为以上4种唇形科药材的品种鉴别提供了一种新方法。
OBJECTIVE: To propose a new method for the rapid identification of four species of Labiatae (Peppermint, Nepeta, Sage and Patchouli) by UV-Vis spectroscopy. Methods: Firstly, the principal component analysis was used to cluster the four medicinal materials and obtain their UV-Vis fingerprint. Then, the artificial neural network technology was used to identify the varieties. Results: The principal component analysis showed that the cumulative confidence of principal component 1 and principal component 2 reached 99%, and the score of all the model samples with principal components 1 and 2 was scored. Good distinction. A three-layer BP artificial neural network model is established by using the characteristic bands sensitive to the four kinds of medicinal herbs as the input of the neural network. The unknown 20 samples were predicted, breed identification accuracy of 100%. Conclusion: This method has a good classification and identification, which provides a new method for the identification of the above four species of Labiatae.