利用一维卷积神经网络模型和原位显微拉曼光谱定量分析牛油果油  

Quantitative Analysis of Avocado Oil Using One-Dimensional Convolutional Neural Network Model and In Situ Micro Raman Spectroscopy

作　　者：张雪松 孙铭思 刘换峥 ZHANG Xuesong;SUN Mingsi;LIU Huanzheng(Jilin Engineering Vocational College,Siping Jilin 136001,China;College of Computer Science and Technology,Jilin University,Changchun Jilin 130012,China;School of Food Science and Engineering,Jilin Agricultural University,Changchun 130118,China)

机构地区：[1]吉林工程职业学院,吉林四平136001 [2]吉林大学计算机科学与技术学院,吉林长春130012 [3]吉林农业大学食品科学与工程学院,吉林长春130118

出　　处：《光散射学报》2024年第4期445-453,共9页The Journal of Light Scattering

基　　金：吉林省教育厅项目(2022ZCY077);吉林省职业技术教育学会项目(2022XHY043)。

摘　　要：牛油果油是从牛油果果肉中提炼的一种新型植物油,由于其价格昂贵和大众认知度有限,市场上很可能会出现制假贩假的现象。为实现快速、无损和高通量的检测需求,本文提出一种利用一维卷积神经网络模型和原位显微拉曼光谱定量分析牛油果油的检测方法。采用菜籽油和葵花油的混合物作为牛油果油掺假的主要成分,并使用原位显微拉曼光谱技术检测了纯植物油和混合油品的光谱,分析并解译了牛油果油的拉曼光谱特征谱峰的化学信息,通过协方差和相关系数遴选了与牛油果油浓度变化存在协同性和相关性的光谱信息,并将其作为网络模型的输入。构建的一维卷积神经网络模型在测试集中预测效果良好,总体的R^(2)>0.915,RMSR<0.0755。基于一维卷积神经网络模型结合原位显微拉曼光谱技术预测牛油果油掺伪浓度的检测方法可行性较好,满足市场应用的检测需求,该成果对于规范国内的牛油果油市场,加快市场监督的职能性管理具有重要的价值。Avocado oil is a new vegetable oil extracted from avocado pulp.Because of its high price and limited public awareness,it is likely to produce and sell fake products.To meet the requirements of rapid,non-destructive,and high-throughput detection,this paper proposes a detection method for avocado oil by using a one-dimensional convolutional neural network model and in-situ micro Raman spectroscopy.The mixture of rapeseed oil and sunflower oil was used as the main component of avocado oil adulteration,and the spectra of pure vegetable oil and mixed oil were detected by in-situ micro Raman spectroscopy technology.The chemical information of the Raman spectrum characteristic peaks of avocado oil was analyzed and interpreted.The spectral information with synergy and correlation with Avocado concentration changes was selected through the covariance difference and correlation coefficient,and it was used as the input of the network model.The one-dimensional convolutional neural network model has a good prediction effect in the test set,with overall R^(2)>0.915 and rmsr<0.0755.The detection method is based on a one-dimensional convolutional neural network model combined with in-situ micro Raman spectroscopy technology to predict the adulteration concentration of avocado oil,which is feasible and meets the detection requirements of market applications.The results have significant value for standardizing the domestic avocado market and accelerating the functional management of market supervision.

关 键 词：原位显微拉曼光谱 一维卷积神经网络 牛油果油 定量分析 

分 类 号：O433[机械工程—光学工程]