光谱学与光谱分析, 2018, 38 (8): 2529, 网络出版: 2018-08-26  

光谱数据融合对绒柄牛肝菌产地溯源研究

Study on the Geographical Traceability of Boletus Tomentipes Using Multi-Spectra Data Fusion
作者单位
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
摘要
由于国内外食品市场准入制度和溯源体系不完善, 销售商乱用虚假标签等现象的发生, 使得食品安全形势愈发严峻。 为了保障野生食用菌的安全性, 保护云南高原特色农业品牌战略, 亟需建立快速准确的产地溯源方法。 通过采集云南及其周边8个产地、 79个绒柄牛肝菌子实体的紫外-可见吸收光谱(UV-Vis)与傅里叶变换红外光谱(FTIR), 采用多元散射校正(MSC)、 标准正态变换(SNV)、 二阶导数(2D)、 平滑(SG)等算法对原始光谱进行预处理。 基于低级与中级数据融合策略, 将预处理后的UV-Vis与FTIR光谱信息进行融合, 结合偏最小二乘判别分析(PLS-DA)与支持向量机(SVM), 建立牛肝菌产地鉴别模型, 确定最佳产地溯源方法。 对光谱融合数据进行系统聚类分析(HCA), 探讨不同产地样品整体化学信息的差异性与相关性。 结果显示: (1)采用MSC+2D和SNV+2D对UV-Vis与FTIR光谱进行预处理, R2Y与Q2最大, 分别为61.58%, 95.09%和50.85%, 82.16%, 表明MSC+2D与SNV+2D为UV-Vis与FTIR光谱的最佳预处理方法; (2)基于UV-Vis, FTIR, 低级与中级数据融合建立的PLS-DA与SVM模型, 样品分类错误总数分别为24, 6, 2, 2和6, 1, 1, 0, 表明数据融合模型分类效果优于单一UV-Vis与FTIR模型; (3)中级数据融合模型中, SVM对所有样品的分类全部正确, PLS-DA的分类错误总数为2, 表明基于SVM的中级数据融合策略分类效果优于PLS-DA; (4)低级和中级数据融合HCA模型, 分别有4和1个样品不能与同一类区域样品聚为一类, 表明中级数据融合优于低级数据融合; 由中级数据融合HCA图可知, 同一产地样品聚类距离小于不同产地之间聚类距离, 表明同一产地样品整体化学成分类较相似, 且同一产地不同采集地点的差异小于不同产地之间的差异。 采用UV-Vis与FTIR光谱中级数据融合策略结合SVM, 能够对不同产地来源牛肝菌样品进行准确鉴别, 为野生食用菌产地溯源研究提供一种新方法。
Abstract
Currently, since the domestic and international food marketing on the safety supervision and traceability system is defective, as well as false labels used by agency, situation on the food safety is becoming more and more serious. In order to enhance food safety, it’s essential to establish a fast and efficient geographical traceability method to protect the agricultural brand of Yunnan plateau. A total of 77 fruit bodies of Boletus tomentipes were collected from 8 geographical origins. Raw of ultraviolet-visible (UV-Vis) and Fourier transform infrared (FTIR) spectra were preprocessed by multiplicative scatter correction (MSC), standard normal variate (SNV), second derivative (2D), Savitzky-Golay (SG) smoothing. Based on pretreatment of UV and FTIR spectra, low-level and mid-level data fusion strategy combined with partial least squares discriminant analysis (PLS-DA) and support vector machine (SVM) were used to identify Boletus in different regions. The results indicated that: (1) that the best pretreatment, was SNV+2D with highest R2Y (61.58%) and Q2 (95.09%) for UV-Vis spectra, and MSC+2D with highest R2Y (50.85%) and Q2 (82.16%) for FTIR spectra; (2) For UV-Vis, FTIR spectra, low-level and mid-level data fusion, the number of error samples in the classification of PLS-DA and SVM analysis were 24, 6, 2, 2, and 6, 1, 1, 0, respectively; (3) In the mid-level data fusion, the best classification of SVM with none error sample was better than that of the PLS-DA with 2 error samples; (4) The classification of HCA analysis in the mid-level data fusion with 4 error samples had the better performance than that in the low-level data fusion with 1 error sample. In addition, HCA analysis of mid-level data fusion showed that the distance of samples collected from same area were longer than that collected from different sites. It indicated that the differences of samples collected from different sites in the same area were less than that collected from different regions. Those results indicated that mid-level data fusion combined with SVM model using UV-Vis and FTIR spectroscopy can accurately identify Boletus collected from different geographical origins. It will provide a new strategy on the research of geographical traceability of wild edible fungus.

张钰, 李杰庆, 李涛, 刘鸿高, 王元忠. 光谱数据融合对绒柄牛肝菌产地溯源研究[J]. 光谱学与光谱分析, 2018, 38(8): 2529. ZHANG Yu, LI Jie-qing, LI Tao, LIU Hong-gao, WANG Yuan-zhong. Study on the Geographical Traceability of Boletus Tomentipes Using Multi-Spectra Data Fusion[J]. Spectroscopy and Spectral Analysis, 2018, 38(8): 2529.

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!