基于高光谱技术和机器学习的半夏伪品鉴别  

Identification of Pinellia ternata Counterfeits Based on Hyperspectral Technology and Machine Learning

作　　者：王昌隆 路绍军 WANG Chang-long;LU Shao-jun(School of Physics and Electronic Information,Yan’an University,Yan’an,Shaanxi 716000)

机构地区：[1]延安大学物理与电子信息学院,陕西延安716000

出　　处：《安徽农业科学》2023年第21期217-220,231,共5页Journal of Anhui Agricultural Sciences

基　　金：延安大学博士启动项目(YDBK2019-55)。

摘　　要：为实现对半夏伪品的鉴别,采用高光谱技术并结合机器学习算法对半夏、水半夏和南星进行研究。首先使用小波变换方法对原始的高光谱数据进行预处理,然后结合主成分分析(PCA)、连续投影算法(SPA)和竞争性自适应重加权采样(CARS)算法在全光谱中提取特征波长,建立了基于全光谱和特征波长的BP神经网络(BP)、支持向量机(SVM)和极限学习机(ELM)的分类判别模型。结果表明,3种特征波长提取方法均能有效提取特征波长,其中基于CARS算法提取的特征波长所建立的分类判别模型效果最佳,而且基于全光谱和CARS算法提取的特征波长建立的BP、SVM、ELM判别模型对训练集和测试集样本的分类识别率均达到了100%。最后,比较了基于全光谱和特征波长建立的ELM模型的运行时间,结果显示基于特征基于波长建立的判别模型运行时间远短于基于全光谱建立的判别模型。该研究为半夏药材的质量控制、伪品鉴别和临床应用奠定基础。In order to identify the counterfeits of Pinellia ternata,hyperspectral technology and machine learning algorithms are used to study Pinellia ternata,Rhizoma Typhonii Flagelliformis and Rhizome Arisaematis.Firstly,the original hyperspectral data was preprocessed by using the wavelet transformation method,and then the characteristic wavelengths were extracted in the full wavelengths by combining principal component analysis(PCA),successive projections algorithm(SPA)and competitive adaptive reweighted sampling(CARS)algorithm,and the classification discriminant model based on the full wavelengths and characteristic wavelengths BP neural network(BP),support vector machine(SVM)and extreme learning machine(ELM)were established.The results showed that the three characteristic wavelengths extraction methods could effectively extract the characteristic wavelengths.Among them,the classification discriminant model based on the feature wavelength extracted by the CARS algorithm had the optimal effects,and the classification recognition rate of the training set and test set samples based on the BP,SVM and ELM discriminant models established by the full wavelengths and the characteristic wavelengths extracted by the CARS algorithm had reached 100%.Finally,the running time of the ELM model based on full wavelengths and characteristic wavelengths was compared,and the results showed that the discriminant model based on characteristic wavelengths ran much shorter than the discriminant model based on full wavelengths.This study laid the foundation for the quality control,counterfeit identification and clinical application of Pinellia ternata medicinal materials.

关 键 词：光谱学 判别模型 鉴别 特征波长 

分 类 号：O433.4[机械工程—光学工程]