多源异构光谱信息融合的食用牛肝菌鉴别方法  

The Identification of Edible Boletus Based on Heterogeneous Multi-Spectral Information Fusion

作　　者：李秀萍 李杰庆[1] 李涛[3] 刘鸿高[1] 王元忠[2] LI Xiu-ping;LI Jie-qing;LI Tao;LIU Hong-gao;WANG Yuan-zhong(College of Agronomy and Biotechnology,Yunnan Agricultural University,Kun ming 650201,China;Institute of Medicinal Plants,Yunnan Academy of Agricultural Sciences,Ku nming 650200,China;College of Resources and Environment,Yuxi Normal University,Yuxi 653100,China)

机构地区：[1]云南农业大学农学与生物技术学院,云南昆明650201 [2]云南省农业科学院药用植物研究所,云南昆明650200 [3]玉溪师范学院资源环境学院,云南玉溪653100

出　　处：《光谱学与光谱分析》2018年第12期3897-3904,共8页Spectroscopy and Spectral Analysis

基　　金：国家自然科学基金项目(31660591;21667031);云南省教育厅科学研究基金项目(2016ZZX106);云南省高校食用菌资源开发与利用重点实验室建设项目资助

摘　　要：牛肝菌营养丰富,味道鲜美,备受各国消费者青睐。因种间差异和环境因素的多层次影响,不同种类及产地牛肝菌品质参差不齐。目前,利益驱动导致商家在牛肝菌销售过程中以次充好、以假乱真的行为扰乱了食用菌市场,不仅给消费者带来健康风险,也制约了牛肝菌的国际化贸易。采用多源异构信息融合策略对牛肝菌种类与产地进行鉴别,以期为追溯食用菌来源以及正确评价其品质提供一种快速有效的解决方法。试验样品灰褐牛肝菌(Boletus griseus)、栗色牛肝菌(B. umbriniporus)、美味牛肝菌(B. edulis)、皱盖疣柄牛肝菌(Leccinum rugosicepes)和绒柄牛肝菌(B. tomentipes)五种牛肝菌科(Boletaceae)真菌子实体采于云南省保山市、昆明市、玉溪市与红河州。采用傅里叶变换红外光谱仪(FTIR)和紫外可见分光光度计(UV-Vis)采集样品信息。Kennard-Stone算法将样品原始数据分为校正集和验证集。校正集基于FTIR、UV-Vis、低级、中级与高级数据融合建立偏最小二乘判别分析(PLS-DA)模型,其中决定系数(R2cal)、预测能力Q2、校正均方根误差(RMSEE)和交叉验证均方差(RMSECV)用来评价模型鲁棒性。研究结果显示:(1)不同种类及产地牛肝菌FTIR和UV-Vis吸收峰的峰位置、峰形和峰数相似,而吸收强度存有差异,表明牛肝菌所含化学成分相似,但含量有一定差别;(2)PLS-DA模型二维散点图可以看出,中级融合比低级融合能更好的鉴别样品种类及产地;(3)各模型中,中级融合模型具有更大的Q2和最小RMSECV,模型鲁棒性最强;(4)验证集样本用来验证模型泛化能力,FTIR、UV-Vis、低级融合、中级融合及高级融合模型样品种类鉴别正确率分别为92. 86%,35. 71%,97. 62%,100%和95. 23%;产地鉴别正确率分别为71. 43%,61. 90%,61. 90%,97. 62%和76. 19%。表明多源异构信息融合在一定程度上优于独立模型,其中,中级数据融合种类鉴别正确率100%,产地鉴�Boletus is rich in nutrition,which is favored b y consumers all over the world.Due to the differences of species and environmen tal factors,the quality of boletus of different species and origin vaires.At present,the shoddy,which undermines the sales of genuine boletus and the mushro om market,not only poses a health risks to consumers,but also restricts the in ternational trade of boletus.In this study,the data fusion stra tegy was used to identify the species and origin of boletus,in order to provide a rapid and effective solution for tracing the source of edible fungi and corr ectly evaluating their quality.The test samples Boletus griseus,B.umbrin iporus,B.edulis,Leccinum rugosicepes and B.tomentipes of five species of boletus fungi fruiting bodies collected from Baoshan,Kunming,Yuxi and Honghe Prefecture of Yunnan province.The chemical information was collected w ith Fourier transform infrared spectroscopy(FT-IR)and UV-Visible spectrophot om eter(UV-Vis).The Kennard-Stone algorithm was used to divide the raw data of samples into calibration sets and validation sets.The calibration set establi shed partial least squares discriminant analysis(PLS-DA)models based on FT-IR,UV-Vis,low-level,mid-level and high-level data fusion.The determin ation coefficients R2 cal,predictive ability Q2,root mean s quare error of estimation(RMSEE)and root mean square error of estimation(RMSE CV)were used to evaluate the robustness of the model.The results showed that:(1)The peak position,peak shape and number of peaks of FT-IR and UV-Vis a bsorption peaks of different species and origin were similar,and there were differences in absorption intensity.This showed that the chemical compositions of boletus were similar,but the content was different.(2)Two-dimensional sc atter plots of PLS-DA model.It can be seen that mid-level fusion is better than low-level fusion to identify sample species and origin.(3)In each model,the mid-level fusion model has a larger Q 2 and a minimum RMSECV,it sh owed that the model has the strongest robustn

关 键 词：牛肝菌 FTIR UV-VIS 多源异构信息融合 种类及产地鉴别 

分 类 号：O567.9[理学—原子与分子物理]