小波矩结合三维荧光光谱简单快速鉴别掺伪芝麻油  

Simple and Quick Identification of Adulterated Sesame Oil by Wavelet Moments and Three-Dimensional Fluorescence Spectra

作　　者：潘钊[1] 李瑞航 吴希军[1] 崔耀耀 PAN Zhao;LI Rui-hang;WU Xi-jun;CUI Yao-yao(Key Lab of Measurement Technology and Instrumentation of Hebei Province,Yanshan University,Qinhuangdao 066004,China;School of Information Science and Engineering,Yanshan University,Qinhuangdao 066004,China)

机构地区：[1]燕山大学河北省测试计量技术及仪器重点实验室,河北秦皇岛066004 [2]燕山大学信息科学与工程学院,河北秦皇岛066004

出　　处：《光谱学与光谱分析》2020年第5期1547-1553,共7页Spectroscopy and Spectral Analysis

基　　金：国家自然科学基金项目(61471312);河北省自然科学基金项目(F2015203072);河北高等学校科学技术研究项目(QN2018071);燕山大学基础研究专项课题(16LGA008)资助。

摘　　要：芝麻油营养丰富,因市场价格较高,掺假现象频出,严重损害了消费者利益和市场的健康发展。因此,研发一种简单快速准确鉴别掺伪芝麻油的方法,对保障消费者权益和市场健康具有重要意义。为此,提出了一种小波矩结合三维荧光光谱掺伪芝麻油鉴别方法。该方法简单快速,计算样本的任一有效特征进行谱系聚类,即可准确鉴别掺伪芝麻油。以43个样本(芝麻油16个,掺伪菜籽油、掺伪大豆油及掺伪玉米油各9个)为研究对象,用FS920荧光光谱仪获得样本的三维荧光光谱。用db2小波将光谱进行多尺度分解(MRSD),用MRSD的一阶离散逼近系数构造小波矩。用前两阶小波矩值W0,0,W1,0,W1,1,W0,1,W2,0,W2,1,W2,2,W1,2,W0,2分别作为特征对样本进行谱系聚类,观察分析聚类结果。结合邓恩分类指数(DVI)进一步分析,研究同阶小波矩分类效果及规律。进而研究各阶小波矩的分类效果及规律。最终确定了用于鉴别掺伪芝麻油的最佳小波矩值。结果表明:MRSD一阶逼近重构光谱可以在保留原光谱的有效特征基础上,大量去除噪声,减少光谱数据量72.4%,增强模型的抗噪稳定性和实时性。利用小波矩前两阶矩值W2,1,W2,2,W1,2,W0,2其一作为分类特征进行谱系聚类,即可鉴别掺伪芝麻油。同阶小波矩(Wp,q)随p值减小q值增大呈现规律性,确定了同阶小波矩的有效矩值及最佳有效矩值。小波矩随着阶数的增加DVI先增后减,最后趋于稳定,确定了各阶小波矩中可用于鉴别掺伪芝麻油的目标矩值W0,q≥2及最佳目标矩值W0,6。小波矩的有效及目标矩值是针对样本分类的有效特征,计算样本的任一有效特征进行谱系聚类,即可实现掺伪芝麻油的鉴别。该研究思路及结论为矩值法应用到三维荧光光谱提供参考。该方法简单快速,可实现在线测量,为质监部门及生产企业提供油品检测和鉴定手段。Sesame oil is rich in nutrients.Due to the high market price,adulteration is frequent,which seriously damages the interests of consumers and the healthy development of the market.Therefore,the development of a fast and accurate method for the identification of adulterated sesame oil(ASO)is of great significance for protecting consumer rights and the market health.To this end,this paper proposed a method for identifying ASO with wavelet moments(WMs)combined with three dimensional fluorescence spectra(3 DFS).This method is simple and rapid,and can effectively identify ASO.In the article,Taking 43 samples(16 sesame oil,9 kinds of rapeseed adulteration sesame oil,soybean adulteration sesame oil and corn adulteration sesame oil,respectively)as research objects.The main research contents and results are as follows:The 3 DFS of the samples were obtained using a FS920 fluorescence spectrometer.Multiresolution signal decomposition(MRSD)was performed on the spectra using db2 wavelets,and then the 3 DFS was reconstructed using the first-order discrete approximation coefficients of MRSD.The first two orders of WMs:W0,0,W1,0,W1,1,W0,1,W2,0,W2,1,W2,2,W1,2,W0,2,were separately used as feature to perform hierarchical clustering(HC)on the samples.Next,combined with Dunn’s cluster validity index(DVI),the classification quality and laws of the same-order and different-order WMs were studied,and the optimal WMs for identifying ASO were determined.Results:MRSD can remove noise and reduce the amount of spectral data by 72.4%on the basis of retaining the effective characteristics of the original spectra.To a certain extent,it can overcome the disadvantages of moment methods that large computational complexity and high-order moments are seriously affected by noise.Using one of W2,1,W2,2,W1,2,W0,2 to perform HC as a feature,the ASO can be easily and quickly identified.The same-order WMs(Wp,q)exhibit regularity as the p decreases q increases,and the effective WMs(EWMs)of the same order were determined.The target moments(TMs)W0,q≥2 and

关 键 词：芝麻油 三维荧光 小波矩 多尺度分析 谱系聚类 邓恩指数 

分 类 号：O657.3[理学—分析化学]