本文選題：銀杏葉 + 近紅外光譜��； 參考：《光譜學(xué)與光譜分析》2017年10期

【摘要】：為考察不同類型光譜信息用于銀杏葉質(zhì)量快速分析的適應(yīng)性,收集了58個(gè)銀杏葉樣品,采用高效液相色譜方法(HPLC)測(cè)定其黃酮及內(nèi)酯類活性成分的含量作為定標(biāo)和檢驗(yàn)樣本的因變量(y)值,測(cè)定各樣品的紫外、近紅外光譜及包含紫外、可見及近紅外信號(hào)的多源復(fù)合光譜信息作為樣本的自變量(x)值;分別采用偏最小二乘回歸(PLSR),以及根據(jù)待測(cè)樣本在自變量空間最近鄰K個(gè)樣本與待測(cè)樣本間的相互關(guān)系去預(yù)測(cè)其因變量值的KNN保形映射(KNN-KSR)方法,建立銀杏葉活性成分的光譜定量分析模型,比較各光譜模型下檢驗(yàn)集樣本實(shí)測(cè)值與模型值的相關(guān)系數(shù)(R)、均方根偏差(RMSEP)、平均相對(duì)誤差(MRE)。結(jié)果表明PLSR方法所建立的三類光譜模型的各項(xiàng)指標(biāo)均不及KNN-KSR方法、且其紫外光譜模型的結(jié)果極差;而采用KNN-KSR方法根據(jù)三類光譜信息預(yù)測(cè)銀杏葉中黃酮、內(nèi)酯類成分時(shí),R基本能達(dá)到0.8、RMSEP分別小于0.05與0.025且其平均相對(duì)誤差均在8%以下。采用KNN-KSR方法根據(jù)紫外、近紅外及多源光譜信息均可實(shí)現(xiàn)對(duì)銀杏葉中四類黃酮醇苷成分及三類內(nèi)酯成分含量的快速分析,突破了現(xiàn)有工作只是基于PLSR方法、根據(jù)近紅外光譜信息對(duì)銀杏葉總黃酮醇苷進(jìn)行定量分析的局限;利用紫外和多源復(fù)合光譜信息及KNN-KSR方法進(jìn)行銀杏葉中黃酮醇苷及內(nèi)酯類成分的快速檢測(cè),為銀杏葉質(zhì)量分析提供了更多的方法和選擇。多源復(fù)合光譜儀具有體積小、成本低,便攜的優(yōu)點(diǎn),非常適合銀杏葉藥材現(xiàn)場(chǎng)采購(gòu)的快速檢測(cè)及后續(xù)產(chǎn)品的質(zhì)量分析與監(jiān)控。
[Abstract]:In order to investigate the adaptability of different spectral information for rapid analysis of ginkgo leaf quality, 58 samples of ginkgo biloba leaf were collected. High performance liquid chromatography (HPLC) was used to determine the contents of flavonoids and lactones as dependent variables of calibration and test samples. The multi-source composite spectral information of visible and near-infrared signals is used as the independent variable of the sample. The partial least squares regression method and the KNN shape preserving mapping KNN-KSRs method are used to predict the dependent variables of K samples in the independent variable space according to the correlation between the K samples and the samples to be tested. The spectral quantitative analysis model of the active components of Ginkgo biloba leaves was established, and the correlation coefficients between the measured values and the model values were compared under each spectral model. The root mean square deviation (RMSEPN) and the mean relative error (MREE) were compared. The results showed that the indexes of the three kinds of spectral models established by PLSR method were not as good as those of KNN-KSR method, and the results of UV spectral model were very poor, and the flavonoids in ginkgo biloba leaves were predicted by KNN-KSR method according to the three kinds of spectral information. The average relative error of RMSEP is less than 0. 05 and 0.025, respectively, and the average relative error is less than 8%. According to the ultraviolet, near infrared and multi-source spectral information, the KNN-KSR method can be used to analyze the contents of tetraflavonol glycosides and three kinds of lactones in Ginkgo biloba leaves, which breaks through the existing work only based on the PLSR method. The limitation of quantitative analysis of total flavonol glycosides in ginkgo biloba leaves based on near infrared spectrum information and the rapid detection of flavonol glycosides and lactones in ginkgo biloba leaves by using ultraviolet and multi-source complex spectral information and KNN-KSR method. It provides more methods and choices for the quality analysis of Ginkgo biloba leaves. Multi-source composite spectrometer has the advantages of small volume, low cost and portable. It is very suitable for the quick detection of field procurement of Ginkgo biloba leaves and the quality analysis and monitoring of subsequent products.
【作者單位】： 華東理工大學(xué)化學(xué)與分子工程學(xué)院;
【基金】：上海市科學(xué)技術(shù)委員會(huì)支撐項(xiàng)目(13401901100)資助
【分類號(hào)】：O433;S792.95
，

本文編號(hào)：1986026