基于近红外技术的初烤烟叶烟碱值预测研究  

Research on Prediction of Nicotine Content of Freshly Flue-cured Tobacco Leaf Based on Handheld Near-infrared Technology

作　　者：周任虎 祝溶 徐杰 丁从凯 李建明 董绍昆 娄勇 杜晓娇 陈颐 Zhou Renhu;Zhu Rong;Xu Jie;Ding Congkai;Li Jianming;Dong Shaokun;Lou Yong;Du Xiaojiao;Chen Yi(Chuxiong Company of Yunnan Provincial Tobacco Corporation,Chuxiong 675000,Yunan,China;Yunnan Mingfan Technology Company Limited,Kunming 650031,Yunan,China;Yunnan Academy of Tobacco Agricultural Sciences,Kunming 650031,Yunan,China)

机构地区：[1]云南省烟草公司楚雄州公司,云南楚雄675000 [2]云南铭帆科技有限公司,云南昆明650031 [3]云南省烟草农业科学研究院,云南昆明650031

出　　处：《山地农业生物学报》2025年第2期56-62,共7页Journal of Mountain Agriculture and Biology

基　　金：中国烟草总公司首批青年科技托举人才项目(2023530401641007);云南省技术创新人才项目(202305AD160036);云南省烟草专卖局重点科技项目(2021530000241035,2022530000241034)。

摘　　要：探究手持式近红外光谱与初烤烟叶烟碱值之间的关系,为烟叶在烘烤过程中不同阶段烟碱值的快速测定提供参考。本研究以初烤烟叶烟碱值预测模型为研究对象,选取542个烟叶样本数据作为训练集,构建了包括主成分回归、偏最小二乘法和Lasso回归等机器学习回归模型,并基于Stacking方法融合为单个预测模型,以136个烟叶样本数据作为测试集评估融合模型的预测效果。融合模型在测试集上的平均绝对误差和决定系数分别为0.218和0.758,且对于烟叶在烘烤过程中不同阶段的烟碱值预测均有效。其中,融合模型对于烟叶类别为“42℃”的烟碱值预测效果最佳,而对于烟叶类别为“鲜烟叶”“38℃”和“46℃”的烟碱值预测效果要优于烟叶类别为“48℃”的烟碱值预测效果。以此方法构建的初烤烟叶烟碱值预测模型在基于手持近红外光谱对烟叶烘烤过程中不同阶段的烟碱值预测方面有着较好的效果。The relationship between handheld near-infrared spectroscopy and the nicotine content of freshly flue-cured tobacco leaves were explored to provide a reference for the rapid determination of nicotine content at different stages of the tobacco curing process.The prediction model of nicotine content of freshly cured tobacco leaves was used as the research object,and 542 tobacco leaf sample data were selected as the training set,and machine learning regression models including principal component regression,partial least squares regression,and Lasso regression were constructed.Then,based on the Stacking method,these models were integrated into a single prediction model.Finally,the prediction performance of the fusion model was evaluated using 136 tobacco leaf sample data as the test set.The results showed that the average absolute error and determination coefficient of the fusion model on the test set were 0.218 and 0.758,respectively,and it was effective in predicting the nicotine content of tobacco leaves at different stages of the curing process.Among them,the fusion model had the best prediction effect on the nicotine content of tobacco leaves with a temperature of“42℃”,while the prediction effect for the nicotine content of tobacco leaves with categories of“fresh tobacco leaves”,“38℃”and“46℃”was better than that for the nicotine content of tobacco leaves with a category of“48℃”.The prediction model of nicotine content in freshly flue-cured tobacco leaves by this method is effective in predicting the nicotine value of tobacco at different stages of the roasting process based on the hand-held near-infrared spectroscopy.

关 键 词：手持近红外光谱 初烤烟叶 烟碱值 模型 机器学习 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]