摘要
为实现采用紫外-可见-近红外光谱技术鉴别掺伪茶油的目的,研究首先通过向茶油中掺入不同比例的葵花籽油、玉米胚芽油和花生油制备掺伪茶油,然后采用自制的透射光谱采集实验平台获得光谱数据,对原始光谱进行预处理后,分别以竞争性自适应重加权算法(CARS)、连续投影算法(SPA)、Boruta算法进行特征波长筛选,最后建立了基于XGBoost的掺伪茶油鉴别模型。研究结果表明,原始光谱经过SG-连续小波变换[CWT(分解尺度25,L5)]预处理和Boruta特征波长筛选后,所建立的XGBoost模型鉴别性能最佳,测试集的准确率、灵敏度和特异性分别达到了98.18%、100.00%和97.62%。通过与常用的支持向量机(SVM)和极限学习机(ELM)模型对比后得到,XGBoost模型的准确率分别提高了3.63%和1.82%,特异性分别提高了4.76%和2.38%。
To realize adulterated camellia oil(CAO)authentication using UV-Vis-NIR spectroscopy,in this study,the samples of adulterated CAO were firstly prepared by mixing different proportions of sunflower oil,corn germ oil and peanut oil,into CAO,respectively.Then,a homemade transmission spectrum acquisition rig was used to obtain their raw spectral data,preprocessed and screened to get the characteristic wavelengths by competitive adaptive re-weighting algorithm(CARS),continuous projection algorithm(SPA),and Boruta algorithm,respectively.Finally,an XGBoost-based authentication model of adulterated CAO was established.The results indicated that the XGBoost model had the best performance after the preprocessing of SG-continuous wavelet transform(CWT,decomposition scalel 25,L5)and Boruta feature wavelength screening.This best model led to the accuracy,sensitivity and specificity of 98.18%,100.00%and 97.62%,respectively.Compared with the commonly used support vector machine(SVM)and extreme learning machine(ELM)models,the accuracy of the XGBoost model was improved by 3.63%and 1.82%,respectively,and the specificity was improved by 4.76%and 2.38%.
作者
龚中良
刘强
李大鹏
文韬
管金伟
易宗霈
申飘
Gong Zhongliang;Liu Qiang;Li Dapeng;Wen Tao;Guan Jinwei;Yi Zongpei;Shen Piao(School of Mechanical and Electrical Engineering,Central South University of Forestry and Technology,Changsha 410004)
出处
《中国粮油学报》
CSCD
北大核心
2023年第8期190-196,共7页
Journal of the Chinese Cereals and Oils Association
基金
湖南省科技计划重点研发项目(2022NK2048)
湖南省教育厅科学项目(18B192,20A515)
湖南省自然科学基金项目(2020JJ4142)
湖南省林业杰青培养科研项目(XLK202108-7)。
