加料回潮工序中烟叶含水率的高光谱在线检测研究  

Hyperspectral Online Detection of Moisture Content on Tobacco Leaves in Feeding Moisture Returning Process

作　　者：马燕玲 赵明霞 李自娟[1] 邢鸿雁[2] 陈娇娇 Ma Yanling;Zhao Mingxia;Li Zijuan;Xing Hongyan;Chen Jiaojiao(Zhangjiakou Cigarette Factor Co.,Ltd.,Zhangjiakou 075000,China;School of Mechanical Engineering,Tianjin University of Science and Technology,Tianjin 300222,China)

机构地区：[1]张家口卷烟厂有限责任公司,河北张家口075000 [2]天津科技大学机械工程学院,天津300222

出　　处：《农机化研究》2025年第5期132-137,172,共7页Journal of Agricultural Mechanization Research

基　　金：国家自然科学基金项目(32101875)。

摘　　要：加料回潮是烟叶制丝过程中的关键工序,其中烟叶含水率的波动偏差直接影响后续卷包机械能否顺利入料加工,故在线精准控制回潮工序中烟叶的含水率至关重要。在线采集162个加料回潮工序中的烟叶样品进行高光谱图像,并使用多元散射校正(MSC)、移动平均(Moving-average)预处理方法对原始光谱进行处理,进一步用主成分分析法(PCA)结合杠杆值法剔除数据集中的奇异值(Novelty),最后采用偏最小二乘回归法(PLSR)创建加料回潮工序中烟叶含水率预测模型。验证结果显示:建立的Moving-average方法预处理的PCA-高杠杆值-PLSR模型最优,其校正集决定系数R_(c)^(2)=0.999,均方根误差RMSEC=0.003,预测集的决定系数R_(p)^(2)=0.999,RMSEP=0.003。研究结论:可以实现烟叶智能监控和快速无损分析,为开发实时检测装备提供理论参考。Nondestructive online detection and control of moisture in tobacco processing is an important procedure in the construction of tobacco intelligent chemical plant.The feeding moisture returning process is the important processes of silk production.The fluctuation deviation of moisture content of tobacco leaves directly affects the smooth feeding and processing of subsequent machinery.It is very important to accurately control moisture content of tobacco leaves in the feeding moisture returning processing online.Firstly,162 tobacco leaf samples were collected online for hyperspectral image acquisition.and the original spectra were processed using Multivariate Scattering Correction(MSC)and Moving Average(Moving-average).Singular values in the dataset were removed using the principal component analysis and leverage value method.Finally,Partial Least Squares Regression(PLSR)was used to create a moisture content prediction model for tobacco leaf in the feeding moisture returning process.The results showed that the Principal Component Analysis(PCA)-High Leverage Value-PLSR model predicted by pretreatment with the Moving-average method had the best prediction performance(the correction set R_(C)^(2)=0.999,and RMSEC=0.003,the prediction set R_(p)^(2)=0.999,and RMSEP=0.003).The results can realize intelligent monitoring and rapid nondestructive analysis of tobacco leaves,and provide theoretical reference for the development of real-time testing equipment.

关 键 词：含水率 加料回潮 高光谱 主成分分析 杠杆值法 偏最小二乘回归 

分 类 号：S123[农业科学—农业基础科学]