| 552 | 0 | 5 |
| 下载次数 | 被引频次 | 阅读次数 |
目的 采用光谱学结合机器学习方法实现菊科红花属植物草红花、鸢尾科番红花属植物西红花(产地伊朗)和藏红花(产地西藏)的种间与种内鉴别。方法 采用拉曼光谱(Raman spectroscopy, RS)与太赫兹(Terahertz, THz)光谱，以伊朗产西红花、甘肃产草红花、西藏产西红花(下文称藏红花)为研究对象，结合主成分分析(Principal component analysis, PCA)、支持向量机(Support-vector machine, SVM)、随机森林(Random forest, RF)等机器学习方法和多层感知机(Multilayer perceptron, MLP)、卷积神经网络(Convolutional neural network, CNN)、长短期记忆递归神经网络(Long short term memory, LSTM)、径向基函数神经网络(Radial Basis Function, RBF)、基于遗传算法的反向传播神经网络(Genetic Algorithm-Back Propagation, GA-BP)等深度学习方法，分析三类样本的THz吸收光谱与RS,实现中草药小样本的快速分类识别。结果 (1)SVM方法对三类样本的RS数据和经PCA降维后的RS数据进行分类精度分别为83.33%和94.44%;MLP对三类样本的原始THz光谱数据和经PCA降维后的THz光谱数据进行分类，准确率分别为94.00%和96.79%;表明PCA对机器学习分类精度提升较为明显，THz光谱用于深度学习结果较RS精度更高。(2)MLP、CNN、RBF、GA-BP、LSTM等神经网络对THz光谱数据的分类准确率高于对RS分类准确率，其中LSTM和RF对太赫兹原始吸收数据分类准确率高达100%,分类精度高、速度快、模型稳定性好。结论 光谱与机器学习方法结合可以实现草红花、西红花和藏红花的种间与种内鉴别，THz光谱鉴别精度较高。
Abstract：Objective To achieve the inter and intraspecific identification of Caohonghua(Carthamus tinctorius L.)from Asteraceae, Xihonghua(Crocus sativus L.)from Iridaceae(originating from Iran)and Zanghonghua(Crocus sativus L.)(originating from Xizang)by using spectroscopic technologies coupled with machine learning methods.Methods Raman spectroscopy(RS)and terahertz(THz)spectral data was used to identify and characterize Xihonghua(Crocus sativus L.)(origin Iran),Caohonghua(Carthamus tinctorius L.)(origin Gansu)and Zanghonghua(Crocus sativus L.)(origin Xizang)by using machine learning methods including Principal Component Analysis(PCA),Support-Vector Machine(SVM),Random Forest(RF)and deep learning methods including Multilayer Perceptron(MLP),Convolutional Neural Network(CNN),Long Short Term Memory(LSTM)and Genetic Algorithm-Back Propagation(GA-BP)to analyze the THz and RS spectra of three types of samples.The rapid classification and recognition of Chinese herbal medicine with small sample size was realized.Results(1)The classification accuracies of the SVM method on the RS data of the three samples before and after dimensionality reduction by PCA were 83.33% and 94.44%,respectively.The MLP classified the THz data before and after dimensionality reduction by PCA with accuracies of 94.00% and 96.79%,respectively.The results indicated that the classification accuracy was further improved by PCA,and THz spectra for deep learning had a higher accuracy than RS spectra.(2)The classification accuracy of neural networks such as MLP,CNN,RBF,GA-BP and LSTM on THz spectral data was higher than that on RS spectra, among which the classification accuracy of LSTM and RF on THz raw data was as high as 100%.Also, these models had high classification accuracy, high speed and good stability.Conclusion The combination of spectroscopic and machine learning methods enables inter and intraspecific identification of Xihonghua(Crocus sativus L.),Caohonghua(Carthamus tinctorius L.)and Zanghonghua(Crocus sativus L.),with high accuracy of THz spectroscopic identification.
[1] 国家药典委员会.中华人民共和国药典：一部[S].北京：中国医药科技出版社，2020:153-1812.
[2] VIDA M,AKHTAR K,EBRAHIM A.Effect of different drying methods on saffron (Crocus sativus l) quality[J].Iran J Chem & Chem Eng Int Engl Ed,2009,31:85-89.
[3] 中国中药协会.西红花质量等级：T/CATCM 002—2018[S].北京：中国标准出版社，2018.
[4] 周桂芬，留永咏，钱晓东，等.基于中国药典与国际标准对西红花生产流通中质量评价方法的改良[J].药物分析杂志，2016,36(5):835-841.
[5] UMBER F,ZAFAR M,ULLAH R,et al.Implication of light and scanning electron microscopy for pollen morphology of selected taxa of family Asteraceae and Brassicaceae[J].Microsc Res Tech,2022,85(1):
[6] D M,TRIFKOVI,et al.Thin-layer chromatography in the authenticity testing of bee-products[J].J Chromatogr B Analyt Technol Biomed Life Sci,2022,1188:123068.
[7] REGUIGUI A,HEIL J,GORAI M,et al.Profile comparison and valorization of Tunisian Salvia aegyptiaca and S.verbenaca aerial part extracts via hyphenated high-performance thin-layer chromatography[J].J Chromatogr A,2022,1673:463057.
[8] GUPTA I,ADIN S N,AQIL M,et al.High-performance thin-layer chromatography method development and validation for quantification of piperine in different extracts of Piper longum L.and its antioxidant activity[J].JPC,2023,36(2):147-155.
[9] KHAIRNAR A,LOHIDASAN S,DUBEY R,et al.Analytical method development and validation for the simultaneous estimation of quercetin,berberine,rutin and curcumin in a polyherbal formulation using high-performance thin-layer chromatography[J].JPC,2023,36(1):63-70.
[10] KIM S,KIM D,JUNG M J,et al.Analysis of environmental organic matters by Ultrahigh-Resolution mass spectrometry-a review on the development of analytical methods[J].Mass Spectrom Rev,2022,41(2):352-369.
[11] TONG Q,ZHANG C,TU Y,et al.Biosynthesis-based spatial metabolome of Salvia miltiorrhiza Bunge by combining metabolomics approaches with mass spectrometry-imaging[J].Talanta,2022,238:123045.
[12] 余志引，朱奇彪.拉曼光谱技术检测真假藏红花[J].现代农业科技，2021(15):223-224,230.
[13] VARLIKLIOZ ER S,EKSI-KOCAK H,YETIM H,et al.Novel spectroscopic method for determination and quantification of saffron adulteration[J].Food Anal Meth,2017,10(5):1547-1555.
[14] DAI H C,GAO Q X,HE L L.Rapid determination of saffron grade and adulteration by thin-layer chromatography coupled with Raman spectroscopy[J].Food Anal Meth,2020,13(11):2128-2137.
[15] 王馨达，宋正勋，李岩，等.基于太赫兹光谱的DNA检测及应用综述[J].激光与光电子学进展，2020,57(17):170003.
[16] 张卫方，杜潮海，刘濮鲲.太赫兹技术在医学成像方面的应用和进展[J].标记免疫分析与临床，2019,26(9):1603-1608.
[17] 马品，杨玉平，范丽洁，等.肝脏和肺组织的太赫兹光谱成像[J].太赫兹科学与电子信息学报，2016,14(6):838.
[18] 马品，杨玉平.运用太赫兹光谱技术检测天麻中的水分含量[J].太赫兹科学与电子信息学报，2017,15(1):26.
[19] 张活.基于太赫兹时域光谱技术的中药检测方法研究[D].西安：西安电子科技大学，2018.
[20] 徐振.基于太赫兹时域光谱技术的川贝母品质检测方法研究[D].南昌：华东交通大学，2022.
[21] 杨玉平，张成，刘海顺，等.两类红花和牛黄的太赫兹光谱法真伪鉴别分析[J].光谱学与光谱分析，2019,39(1):45.
[22] BAI C Y,NIU Q,YANG G,et al.Surface-enhanced Raman spectroscopy substrate based on silver/titanium dioxide nanocomposites and demonstration on rhodamine 6G and severe acute respiratory syndrome coronavirus 2 spike-1 protein[J].Spectrosc Lett,2022,55(6):401-413.
[23] TARANTILIS P A,BELJEBBAR A,MANFAIT M,et al.FT-IR,FT-Raman spectroscopic study of carotenoids from saffron (Crocus sativus L.) and some derivatives[J].Spectrochim Acta Part A Mol Biomol Spectrosc,1998,54(4):651-657.
[24] SCHULZ H,BARANSKA M,BARANSKI R.Potential of NIR-FT-Raman spectroscopy in natural carotenoid analysis[J].Biopolymers,2005,77(4):212-221.
[25] PRATA J C,DA COSTA J P,FERNANDES A J S,et al.Selection of microplastics by Nile Red staining increases environmental sample throughput by micro-Raman spectroscopy[J].Sci Total Environ,2021,783:146979.
[26] MARY S J J,SIDDIQUE M U M,PRADHAN S,et al.Quantum chemical insight into molecular structure,NBO analysis of the hydrogen-bonded interactions,spectroscopic (FT-IR,FT-Raman),drug likeness and molecular docking of the novel anti COVID-19 molecule 2-[(4,6-diaminopyrimidin-2-yl)sulfanyl]-N-(4-fluorophenyl)acetamide - dimer[J].Spectrochim Acta A Mol Biomol Spectrosc,2021,244:118825.
[27] QU F,FU H Y,LI Y Q,et al.Temperature effect on electronic and vibrational properties of β-carotene aggregates in aqueous ethanol solution[J].Dyes Pigm,2019,166:323-329.
基本信息:
DOI：10.13193/j.issn.1673-7717.2025.06.006
中图分类号:TP18;R282.5
引用信息:
[1]苏康慧,王超琪,张谷令等.基于深度学习的拉曼/太赫兹光谱对藏红花的分类识别[J].中华中医药学刊,2025,43(06):36-39+262-266.DOI:10.13193/j.issn.1673-7717.2025.06.006.
基金信息:
国家重点基础研究发展计划项目(2017YFB00405402,2020YFB2009303); 国家自然科学基金项目(62075248)