高光谱结合人工神经网络对丹参药材的质量评价  

Quality Evaluation of Salvia Miltiorrhiza by Hyperspectrum Analysis Combined with Artificial Neural Network

作　　者：李子涵 焦龙[1] 孙妩娟[1] 李红 李栋 娄俊豪 LI Zihan;JIAO Long;SUN Wujuan;LI Hong;LI Dong;LOU Junhao(School of Chemistry and Chemical Engineering,Xi'an Shiyou University,Xi'an 710065,China)

机构地区：[1]西安石油大学化学化工学院,陕西西安710065

出　　处：《化工技术与开发》2024年第3期64-67,25,共5页Technology & Development of Chemical Industry

基　　金：国家自然科学基金项目(211723003);陕西省教育厅青年创新团队建设科研计划项目(21JP097,22JP064);大学生创新创业训练计划项目(202210700010);川庆钻探公司-西安石油大学致密油气藏勘探开发研究中心科技项目(CQXA-2023-05);西安石油大学科研创新团队项目(2019QNKYCXTD17)。

摘　　要：采用高光谱结合人工神经网络,建立了不同来源的丹参药材的鉴别方法。采集了8类不同来源丹参药材样品的高光谱数据,用Savitzky-Golay平滑滤波方法对高光谱数据进行预处理,再采用反向传播-人工神经网络方法建立分类模型。结果显示,经Savitzky-Golay平滑滤波光谱预处理后,建立的模型对测试集的分类准确率达到98.75%。研究表明,高光谱结合人工神经网络方法是一种很有前景的丹参药材的质量评价方法。A method for identifying Salvia miltiorrhiza from different sources was established by hyperspectral analysis combined with artificial neural networks.Hyperspectral data from 8 different sources of Salvia miltiorrhiza medicinal samples was collected,the hyperspectral data was preprocessed by Savitzky Golay smoothing filtering method,and then a classification model was established by back propagation-artificial neural network method.The results showed that after Savitzky Golay smooth filtering spectral preprocessing,the established model achieved a classification accuracy of 98.75%for the test set.Research showed that the combination of hyperspectral and artificial neural network methods was a promising method for evaluating the quality of Salvia miltiorrhiza medicinal herbs.

关 键 词：高光谱 丹参 人工神经网络 中药材质量评价 

分 类 号：O657.3[理学—分析化学] R282.5[理学—化学]