Page 30 - 《中国药房》2022年22期
P. 30
基于机器视觉系统的姜炭炮制程度判别及颜色-成分相关性
分析
Δ
1 #
1
1
1
1*
1
2
张一凡 ,周苏娟 ,孟 江 ,左 蓉 ,林华坚 ,孙 悦 ,王淑美 [1.广东药科大学中药学院/国家中医药管理局
中药数字化质量评价技术重点研究室/广东高校(省)中药质量工程技术研究中心,广州 510006;2.广东药科大
学信息工程学院,广州 510006]
中图分类号 R917;R283 文献标志码 A 文章编号 1001-0408(2022)22-2712-07
DOI 10.6039/j.issn.1001-0408.2022.22.05
摘 要 目的 基于机器视觉系统探讨姜炭炮制程度判别及颜色-成分含量相关性,为姜炭炮制程度控制和质量评价提供参考。
方法 采用高效液相色谱法测定干姜及其不同炮制程度姜炭饮片中6-姜酚、8-姜酚、10-姜酚、6-姜烯酚、姜酮5种成分的含量;采用
* *
机器视觉技术获得饮片图像并提取饮片RGB、L a b 、HSV 3个不同颜色空间的颜色特征,采用机器学习,如主成分分析(PCA)、线
*
性判别分析(LDA)、偏最小二乘法-判别分析(PLS-DA)和支持向量机(SVM)等方法对姜炭不同炮制程度姜炭建立定性判别模型,
将颜色特征值与测得的5种成分含量进行相关性分析,建立颜色-成分的定量预测模型。结果 随着炮制程度的加深,姜酮在炮制
后产生且含量先增加后降低,标炭饮片中含量最高;6-姜酚、8-姜酚和10-姜酚的含量逐渐降低;6-姜烯酚含量先增加后降低。基于
饮片图像颜色客观量化结合有监督判别模式识别方法的 LDA 和 SVM 建立的定性判别模型,其交叉验证训练预测准确率达到
100%,外部验证准确率达到95.83%。基于饮片图像颜色客观量化结合SVM建立5种成分含量预测模型,RPD值均大于2,各成分
的R P与R C值中姜酮为0.633 9与0.683 3,其他成分值均大于0.75,说明SVM对除姜酮以外的4种成分均有较好的预测能力。结论
2
2
基于机器视觉系统对姜炭炮制程度的判别和含量预测模型效果良好,可为姜炭饮片的质量控制和炮制程度的判断提供参考。
关键词 姜炭;机器视觉;机器学习;质量评价;炮制程度
Discrimination of processing degree of Zingiber officinale charcoal and analysis of the correlation between
color and component based on machine vision system
1
ZHANG Yifan ,ZHOU Sujuan ,MENG Jiang ,ZUO Rong ,LIN Huajian ,SUN Yue ,WANG Shumei [1. School
2
1
1
1
1
1
of Traditional Chinese Medicine, Guangdong Pharmaceutical University/Key Laboratory of Digital Quality
Evaluation of Chinese Materia Medica, State Administration of Traditional Chinese Medicine/Engineering
Technology Research Center for Chinese Materia Medica Quality of Guangdong Universities and Colleges
(Province), Guangzhou 510006, China;2. College of Medical Information Engineering, Guangdong
Pharmaceutical University, Guangzhou 510006, China]
ABSTRACT OBJECTIVE To explore the discrimination of processing degree of Zingiber officinale charcoal and the correlation
between color and component content based on machine vision system, and provide reference for quality evaluation and processing
degree control of Z. officinale charcoal. METHODS High-performance liquid chromatography method was used to determine the
contents of 5 components in Z. officinale charcoal and its different processed products, such as 6-gingerol, 8-gingerol, 10-gingerol,
6-shogaol, gingerone. Machine vision system was used to obtain the image of the decoction pieces and extract the color features of
the decoction pieces in RGB, L a b and HSV color spaces. Machine learning methods, such as principal component analysis
* *
*
(PCA), linear discriminant analysis (LDA), partial least squares-discriminant analysis (PLS-DA) and support vector machine
(SVM), were used to establish qualitative identification model for Z. officinale charcoal processed products of different processing
degree. The correlation between the color eigenvalues and the contents of measured 5 components were analyzed, and the color-
component content prediction model was established.
Δ 基金项目 国家自然科学基金资助项目(No.81873006)
RESULTS With the deepening of processing, gingerone was
*第一作者 硕士。研究方向:中药炮制原理与质量标准。E-mail:
1416812771@qq.com produced after processing and the content firstly increased and
# 通信作者 教授,硕士生导师,博士。研究方向:中药炮制及饮片 then decreased, and the content of gingerone in standard
质量控制。电话:020-39352175。E-mail:jiangmeng666@126.com carbon was the highest; the contents of 6-gingerol, 8-gingerol
· 2712 · China Pharmacy 2022 Vol. 33 No. 22 中国药房 2022年第33卷第22期