近红外光谱技术结合教与学算法优化极限学习机实现烤烟等级判定  

Grade Determination of Flue-cured Tobacco by Near Infrared Spectroscopy Combined with Teaching-learning-based Optimization Algorithm Optimized Extreme Learning Machine

作　　者：沈欢超 耿莹蕊 倪鸿飞 王辉[3] 吴继忠[3] 廖付 陈勇[1] 刘雪松[1] SHEN Huan-chao;GENG Ying-rui;NI Hong-fei;WANG Hui;WU Ji-zhong;LIAO Fu;CHEN Yong;LIU Xue-song(College of Pharmaceutical Sciences,Zhejiang University,Hangzhou 310058,China;Innovation Institute for Artificial Intelligence in Medicine of Zhejiang University,Hangzhou 310018,China;Technology Center,China Tobacco Zhejiang Industrial Co.,Ltd.,Hangzhou 310008,China)

机构地区：[1]浙江大学药学院,浙江杭州310058 [2]浙江大学智能创新药物研究院,浙江杭州310018 [3]浙江中烟工业有限责任公司技术中心,浙江杭州310008

出　　处：《分析测试学报》2022年第7期1052-1057,共6页Journal of Instrumental Analysis

基　　金：浙江大学-浙江中烟联合实验室项目资助。

摘　　要：该研究基于近红外光谱(NIRs)技术,以2016~2018年来自13个省份的937个烟叶样本为研究对象,比较了竞争性自适应重加权采样方法(CARS)、蒙特卡洛无信息变量消除法(MC-UVE)以及随机青蛙算法(RF)3种变量筛选方法的极限学习机(ELM)模型效果,与常规判别方法偏最小二乘判别分析(PLS-DA)比较,验证了ELM模型的优势。并通过教与学优化(TLBO)算法对ELM模型进行优化,建立烤烟样本的等级判定模型。结果表明,验证集的分类正确率达到90.16%,测试集的外部验证表现良好,TLBO-ELM模型收敛速度快,泛化能力强,可应用于烤烟等级判定。近红外光谱技术结合教与学算法优化极限学习机为智能化实现烟叶等级判定提供了一种新方法。The quality evaluation on tobacco is an important work as it is a high-value attribute product.Therefore,it is of a certain application value to ultilize intelligent means for efficient classification of tobacco.Based on near infrared spectroscopy(NIRs),937 tobacco samples from 13 provinces from 2016 to 2018 were used to compare the extreme learning machine(ELM)model effects of three variable screening methods,including competitive adaptive reweighted sampling(CARS)method,Monte Carlo uninformed variable elimination(MC-UVE)method and random frog(RF)algorithm.Compared with partial least squares-discriminant analysis(PLS-DA),the advantages of ELM model were verified.The ELM model was optimized by teaching-learning-based optimization(TLBO)algorithm,thus a TLBO-ELM classification model for flue-cured tobacco samples was established.Results showed that the classification accuracy of the validation set was 90.16%.The external verification effect of the testing set was satisfactory,and the TLBO-ELM model had fast convergence speed and strong generalization ability,which could be applied to the classification of flue-cured tobacco.NIRs combined with TLBO to optimize ELM provides a new idea for intelligent tobacco classification.

关 键 词：近红外光谱技术 教与学优化算法 极限学习机 烟叶 等级判定 

分 类 号：O657.3[理学—分析化学] F768.29[理学—化学]