基于NIR技术和ELM的烤烟烟叶自动分级  

Automatic grading of flue-cured tobacco leaves based on NIR technology and extreme learning machine algorithm

作　　者：宾俊[1] 周冀衡[1] 范伟[1] 李鑫[1] 梁逸曾[2] 肖志新[3] 李春顺[4] 

机构地区：[1]湖南农业大学,生物科学技术学院,湖南长沙410128 [2]中南大学,化学化工学院,湖南长沙410083 [3]云南省烟草公司保山市公司,云南678000 [4]江苏中烟工业有限责任公司,江苏南京210019

出　　处：《中国烟草学报》2017年第2期60-68,共9页Acta Tabacaria Sinica

基　　金：国家自然科学基金资助项目(No.21275164);湖南省研究生科研创新资助项目(No.CX2015B237)

摘　　要：为解决初烤烟叶收购中人工分级主观因素影响较大的问题,提出了一种基于近红外(NIR)光谱技术结合极限学习机(ELM)算法自动鉴别烟叶等级的方法。文章首次提出基于品质相似、价格接近原则的烟叶收购分组方法,通过交互检验优化ELM分组、分级模型的隐节点数,并与K最近邻法(KNN)、支持向量机(SVM)和随机森林(RF)等多分类算法进行了比较。结果表明:ELM分类模型参数自动优化、训练时间短、稳定性和预测能力较好,2014年(数据集A)、2015年(数据集B)烟叶收购国标样本上、中、下等烟外部预测分组正确率分别为95.77%和94.23%,数据集A和B的上、中、下等烟各组样本外部预测分级正确率分别为85.71%、86.67%、100%和100%、92.86%、92.86%。因此,采用NIR技术结合ELM能准确鉴别初烤烟叶等级,可为烤烟烟叶收购质量等级评价提供一种新技术。In order to minimize the influence of artificial experience on flue-cured tobacco leaf grading in purchasing process, a rapid grading method using near-infrared （NIR） spectroscopy combined with extreme learning machine （ELM） algorithm was proposed. A grouping method based on principle of similar quality and close price of flue-cured tobacco leaves was put forward. Cross validation was used to optimize the number of hidden nodes of ELM. The method was compared with commonly used multi-class classification algorithms, including K nearest neighbor （KNN）, support vector machine （SVM）, and random forest （RF） algorithm. Results showed that ELM classification model was superior to other methods with automatic optimization parameters, short training time, and high stability and predictability. The classification prediction accuracy of tobacco dataset A and B into high, medium, and low groups was 95.77% and 94.23%, respectively. Furthermore, classification accuracy of subdividing high, medium, and low groups of tobacco prediction samples A was 85.71%, 86.67%, and 100%, respectively, and subdivision accuracy of tobacco prediction samples B was 100%, 92.86% and 92.86%, respectively. Therefore, application of NIR technology combined with ELM could accurately determine flue-cured tobacco leaf grade, providing a promising tool for quality evaluation in flue-cured tobacco leaf purchasing process.

关 键 词：烟叶分级 近红外光谱 极限学习机 分类模型 多分类算法 

分 类 号：TS442[农业科学—烟草工业]