Research On Variation Of Nutrient Components And Classification In Lily Bulbs Based On Raman Spectroscopy

Raman spectroscopy is a kind of vibrational spectroscopy based on inelastic scattering,which is specific to different chemical bonds or groups,so it can identify the molecular structure characteristics and types of the sample at the molecular level.It has the advantages of high speed and accuracy,and can be used for nondestructive testing,so Raman spectroscopy is being increasingly used in biochemical testing and other fields.Lily is a typical crop used for both medicine and food,which is rich in nutrients and has an important healthcare function.Based on the Raman spectroscopy technique,we collected the Raman spectra of the three most widely distributed lilies（including Lanzhou lily,Yixing lily and Longya lily）in China in this paper,then Raman spectra of different Lanzhou lily samples were analyzed.From the spectroscopic point of view,the main components in the scales of Lanzhou lily are quite different:the intensity of some characteristic peaks at 479 cm^-1,870 cm^-1 and 942 cm^-1 is very high in the matrix of the inner and outer scales,indicating the high content of starch and fructose.The intensity of characteristic peak at 1606 cm^-1 representing ferulic acid in the spectra of outer scales’epidermis and inner scales’matrix is very high.The characteristic peaks of phospholipids and multiple amino acids are found in the spectra of all matrix samples.In addition,the Raman spectra of Lanzhou lilies with different storage periods are consistent with their own physiological characteristics.Lanzhou lilies can be marketed for most of the year,but the quality of Lanzhou lilies produced during the intensive harvesting period is the best.The"single-head"Lanzhou lilies with complete shape and no"split"phenomenon has the highest carbohydrate and amino acid content.These conclusions provide a lot of important references for scientific evaluation of the quality of Lanzhou lily.In order to prove the above conclusions,a method for quantitative estimation of component content based on Raman spectroscopy was proposed.The results of quantitative estimation for starch and sucrose content in lily are consistent with the results of the previous spectroscopy analysis.The results of measurement and analysis show that Raman spectroscopy can conduct in-depth and detailed research on the nutrient content and change rules of Lanzhou lily and it is a scientific and efficient detection method.There are many varieties of lilies in China,but the quality of lilies on the market is uneven and the origin is vague,which makes it difficult to fully exert their health benefits.In this paper,Raman spectroscopy and machine learning algorithms are combined to establish classification models of three kinds of lilies.Three methods including manual extraction,principal component analysis and t-distribution stochastic neighbor embedding are used to extract the characteristics of spectral data.The extracted data are applied to algorithms such as support vector machine,decision trees,and random forest.The results demonstrate that these classification models all show ideal classification accuracy on the same test set.Among them,the accuracy of the classification model based on t-SNE and SVM is 93.7%.Based on PCA and decision tree algorithm,the model accuracy of the classification model can reach 91.7%,and the accuracy is as high as 95.8%based on PCA and RF.Experimental results show that Raman spectroscopy combined with machine learning algorithms can quickly identify the origin of lilies.