本文選題：近紅外光譜　切入點：完整煙葉　出處：《光譜學與光譜分析》2017年10期

【摘要】：為了探討近紅外光譜分析技術檢測完整煙葉化學成分的可行性,利用近紅外光譜分析技術,對初烤完整煙葉的光譜采集方式及總植物堿定量分析建模方法進行了研究。以云南省昆明市不同鄉(xiāng)鎮(zhèn)、不同品種的初烤煙葉為研究對象,分別采用煙葉的葉尖、葉中、葉基光譜及其平均光譜建立初烤完整煙葉總植物堿近紅外偏最小二乘法(PLS)定量分析模型以選擇出代表完整煙葉信息的建模光譜;分別用KS和SPXY方法對樣品的校正集進行選擇,采用向后區(qū)間偏最小二乘法(BiPLS)、無信息變量消除法(UVE)、競爭適應性重加權采樣法(CARS)等選擇特征變量,對模型進一步優(yōu)化。研究結果表明,采用葉尖、葉中、葉基3個部位的平均光譜建立的模型相比單獨每個部位光譜所建立模型的預測精度提高了8.5%~9.5%,與全光譜建模相比,用KS-BiPLS建立模型能明顯改善模型的預測能力,模型的預測精度約提高了10%,模型的校正集決定系數(shù)和均方根誤差分別為0.917 4和0.226 1,檢驗集決定系數(shù)和預測均方根誤差分別為0.902 0和0.2007。本研究方法適用于完整的初烤煙葉,無需對樣品進行預處理,對于大量的初烤煙葉,能夠快速、無損測定煙葉總植物堿含量,可以節(jié)省大量的時間。同時,該研究為初烤煙葉分級、提高原料的品質(zhì)提供技術支持,也將為卷煙生產(chǎn)的過程控制提供科學依據(jù)。
[Abstract]:In order to explore the feasibility of detecting the chemical components of intact tobacco leaves by near infrared spectroscopy (NIR), the spectral acquisition method and the quantitative analysis modeling method of total plant alkaloids were studied by using Near-infrared spectroscopy (NIR).The first baked tobacco leaves of different villages and towns in Kunming, Yunnan Province, were studied. The leaf tip and middle leaf of tobacco leaf were used respectively.The leaf base spectrum and its average spectrum were used to establish a quantitative analysis model of total alkaloids of the whole tobacco leaves, and to select the modeling spectrum representing the complete tobacco leaf information, and the calibration sets of the samples were selected by using KS and SPXY methods, respectively.The model is further optimized by using the backward interval partial least square method (BiPLS), the elimination of no information variables (UVEV), and the competitive adaptive re-weighted sampling method (CARSs).The results show that the prediction accuracy of the model based on the average spectra of the three parts of leaf tip, leaf and leaf base is 8.59.5than that of the model established by each part alone, which is compared with the full-spectrum model.The prediction ability of the model can be improved obviously by using KS-BiPLS. The prediction accuracy of the model is improved by about 10%. The calibration set decision coefficient and root mean square error of the model are 0.917 4 and 0.226 1 respectively, and the determination coefficient of test set and the prediction root mean square error are 0.902 0 and 0.2007 respectively.This method is suitable for the whole freshly baked tobacco leaves without pretreatment. For a large number of freshly baked tobacco leaves, it can be used to determine the total plant alkali content of tobacco leaves quickly and without damage, and can save a lot of time.At the same time, the research will provide technical support for the classification of freshly baked tobacco leaves and improve the quality of raw materials, as well as provide scientific basis for the process control of cigarette production.
【作者單位】： 中國農(nóng)業(yè)大學現(xiàn)代精細農(nóng)業(yè)系統(tǒng)集成研究教育部重點實驗室;云南省煙草農(nóng)業(yè)科學研究院;
【基金】：國家自然科學基金項目(61144012) 中國煙草總公司云南省公司項目(2015YN03)資助
【分類號】：O657.33;TS411
，

本文編號：1718633