基于近红外光谱技术和变量筛选-偏最小二乘判别分析方法的铁皮石斛产地无损溯源  

Origin Traceability of Dendrobium candidum with Near Infrared Spectroscopy and Variable Selection-Partial Least Squares Discriminant Analysis

作　　者：柳薇 邱熙文 蒋立文 范伟 李跑[2,3] 郑郁 LIU Wei;QIU Xi-wen;JIANG Li-wen;FAN Wei;LI Pao;ZHENG Yu(School of Medicine,Hunan Normal University,Changsha 410013,China;College of Food Science and Technology,Hunan Agricultural University,Changsha 410128,China;Hunan Agricultural Product Processing Institute,Hunan Academy of Agricultural Sciences,Changsha 410125,China)

机构地区：[1]湖南师范大学医学院,湖南长沙410013 [2]湖南农业大学食品科学技术学院,湖南长沙410128 [3]湖南省农业科学院湖南省农产品加工研究所,湖南长沙410125

出　　处：《分析测试学报》2025年第2期246-252,共7页Journal of Instrumental Analysis

基　　金：国家自然科学基金项目(31601551);湖南省自然科学基金面上项目(2023JJ30290);广东省粤北食药资源利用与保护重点实验室开放基金资助项目(FMR2023012Z)。

摘　　要：基于近红外(NIR)光谱技术与变量筛选-偏最小二乘判别分析(PLS-DA)方法,建立了一种铁皮石斛产地的无损溯源方法。利用光栅型便携式NIR光谱仪采集了3个积分时间(25、45、65 ms)下3个产地共900份铁皮石斛枫斗的光谱。采用光谱预处理方法消除光谱中的干扰;以主成分分析(PCA)、PLS-DA方法建立了铁皮石斛产地的鉴别模型;通过竞争性自适应重加权采样法(CARS)、蒙特卡罗-无信息变量消除法(MCUVE)及连续投影算法(SPA)筛选获得特征变量进一步提升PLS-DA模型鉴别准确性;此外,首次探究了积分时间对铁皮石斛产地溯源模型的影响。结果表明:无论是原始光谱还是优化预处理后的光谱,PCA方法均无法实现铁皮石斛产地的准确鉴别;45 ms和65 ms积分时间的PLS-DA模型可以实现石斛产地的100%鉴别分析;CARS和MC-UVE模型显著优于SPA模型,可在获得较少变量数的前提下实现石斛产地的100%鉴别分析。以上结果表明,基于便携式NIR光谱技术与变量筛选-PLS-DA方法可实现对铁皮石斛产地的准确鉴别,为中药材的质量控制研究提供了新方向。A non-destructive origin traceability method for Dendrobium candidum was established based on near infrared(NIR)spectroscopy technology and variable selection-partial least squares dis⁃criminant analysis(PLS-DA)method.900 spectra of Dendrobium candidum Fengdou from three ori⁃gins were collected using a grating type portable NIR spectrometer at three integration times(25 ms,45 ms,65 ms).The interferences in the spectra were eliminated using spectral pretreatment meth⁃ods.A model for identifying the origin of Dendrobium candidum was established using principal com⁃ponent analysis(PCA)and PLS-DA methods.The characteristic variables were selected by competi⁃tive adaptive reweighted sampling(CARS),Monte Carlo-uninformative variable elimination(MC-UVE)and successive projections algorithm(SPA)to further improve the accuracy of the PLS-DA model.In addition,it is also the first study to investigate the impact of integration time on the mod⁃els.The results showed that the PCA method cannot achieve the accurate identification of Dendrobi⁃um candidum origins with the original and the optimized preprocessed spectra.Using the integration times of 45 ms and 65 ms,the PLS-DA model with the optimized spectral pretreatments can achieve 100%identification of Dendrobium candidum origins.The CARS and MC-UVE models were significantly better than the SPA model,which can be used to achieve 100%identification of Dendrobium candidum origins with fewer variables.The research indicates that the portable NIR spectroscopy technology combined with variable selection-PLS-DA method can achieve accurate identification of Dendrobium candidum origins,providing a new direction for the quality control research of tradition⁃al Chinese medicinal materials.

关 键 词：近红外光谱技术 铁皮石斛 产地溯源 变量筛选方法 偏最小二乘判别分析 

分 类 号：O657.3[理学—分析化学] R282[理学—化学]