基于仿生嗅觉的中药材指纹图谱建立与鉴别方法的研究

发布时间：2018-04-30 23:17

本文选题：仿生嗅觉 + 电子鼻　；参考：《广东工业大学》2012年硕士论文

【摘要】：我国中药材资源丰富,对人类医学发展和促进人体健康发挥着巨大的作用。但由于药材种类繁多,市场上出现大量的假冒伪劣产品,就连普通的中药材都出现了大量的混淆品,严重影响了中医药的发展。因此,中药材的品质判定一直是人们研究的热点,其中产地因素又是评判中药材品质的重要标准之一。但是长期以来,国内外对于中药材品质的评定,普遍采用的是感官评审法,然而,感官评审法往往要受诸多因素的影响。这就对中药材品质的检测和评判提出了更高的要求,要求其更加科学和规范。气味在中药材品质分析中占有重要地位,仿生嗅觉技术模拟了人类嗅觉的原理,通过检测中药材挥发性气味的整体信息来自动完成对气味的辨识。目前,国内外关于将仿生嗅觉技术运用于中药材领域的研究报道还相对比较少,因此我们拟通过仿生嗅觉技术来检测中药材挥发出的综合信息,建立一种评价中药材的新技术。研究以姜科、伞形科和菊科三种典型科属的中药材作为研究对象,通过PEN3电子鼻检测并提取其各特征值,生成高维的特征向量。然后采用主成分分析法提取其相应的主成分分量,构成模式识别的输入。结合聚类分析和BP神经网络两种模式识别方法来实现对不同产地以及易混淆中药材的判别与鉴定,最后建立适量的中药材气味指纹图谱库。聚类分析结果显示能够正确的对各待测样品进行归类。采用BP神经网络的方法对不同产地白术训练集的回判正确率均为100%,误判的待测样本只发生在安徽白术,其判断正确率为86.67%；对易混淆的三组药材训练集的正确率均为100%,只有砂仁发生误判,其判断正确率为93.33%。对两种模式识别方法的优劣进行分析和对比,得出结论：聚类分析由于其算法简单,能够快速的对样品进行分类,但如果使用复杂的距离相似度度量时,计算复杂度会提高,使其不再具备快速简便的优点；BP神经网络具有高度的非线性,在理论上可以逼近任意曲面,但是计算量较大,计算复杂度也较高。本文实验最合适的方法是聚类分析。最后利用基于统计特征(均值、方差、峰值)的3种方法来构建样品的指纹图谱库,结果发现,指纹图谱曲线具有较高的区分度,并且各待测样本的指纹图谱曲线都能够与库中相应的指纹图谱曲线基本相吻合。结果显示,采用PEN3电子鼻能够正确实现中药材的分类鉴别和指纹图谱库的构建。
[Abstract]:Chinese medicinal materials are rich in resources and play a great role in the development of human medicine and the promotion of human health. However, because of the variety of medicinal materials, a large number of fake and shoddy products appear in the market, even ordinary Chinese medicinal materials appear a large number of confounding products, which seriously affect the development of traditional Chinese medicine. Therefore, the quality evaluation of Chinese medicinal materials has been the focus of research, among which the origin factor is one of the important criteria to judge the quality of Chinese medicinal materials. However, for a long time, the sensory evaluation method is widely used in the quality evaluation of Chinese medicinal materials at home and abroad. However, the sensory evaluation method is often affected by many factors. Therefore, the quality of traditional Chinese medicine is required to be more scientific and standardized. Smell plays an important role in the quality analysis of Chinese medicinal materials. Bionic olfactory technology simulates the principle of human olfaction and automatically realizes the identification of smell by detecting the whole information of volatile odors of Chinese medicinal materials. At present, there are relatively few reports on the application of bionic olfactory technology in the field of Chinese medicinal materials at home and abroad. Therefore, we intend to use bionic olfactory technology to detect the volatile information of Chinese medicinal materials and establish a new technology to evaluate Chinese medicinal materials. In this study, the traditional Chinese medicines of three typical families and genera of ginger, umbrella and Compositae were used as the research objects. The PEN3 electronic nose was used to detect and extract each characteristic value and to generate the high dimensional characteristic vector. Then principal component analysis (PCA) is used to extract the corresponding principal components to form the input of pattern recognition. Cluster analysis and BP neural network were used to identify and distinguish Chinese medicinal materials from different habitats and easily confused. Finally, a proper amount of odor fingerprint library was established. Cluster analysis results show that each sample can be correctly classified. Using BP neural network method, the correct rate of correct judgment for the training set of Atractylodes macrocephala from different areas is 100, and the samples to be misjudged only occur in Anhui Atractylodes macrocephala. The correct rate of judgment was 86.67, and the correct rate of the three sets of medicine training was 100, only Amomum villosum was misjudged, and the correct rate of judgment was 93.33. The advantages and disadvantages of the two pattern recognition methods are analyzed and compared. It is concluded that the clustering analysis can quickly classify samples because of its simple algorithm, but the computational complexity will be increased if the complex distance similarity measure is used. The BP neural network has a high degree of nonlinearity and can approach any surface in theory, but it has a large amount of computation and a high computational complexity. The most suitable method for this experiment is clustering analysis. Finally, three methods based on statistical features (mean, variance, peak) are used to construct the fingerprint database of samples. The results show that the fingerprint curve has a high degree of differentiation. And the fingerprint curve of each sample can be basically consistent with the corresponding fingerprint curve in the database. The results showed that the classification and identification of Chinese medicinal materials and the construction of fingerprint library could be correctly realized by using PEN3 electronic nose.
【学位授予单位】：广东工业大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH788.2

【参考文献】