浙江农业科学 ›› 2026, Vol. 67 ›› Issue (3): 608-616.DOI: 10.16178/j.issn.0528-9017.20250019

• 经济作物 • 上一篇    下一篇

基于K⁃means聚类的烟叶物理特性分型方法

许强1(), 苏雪苗2, 苗晨琳1, 张艺帆1, 杨月1, 陈超3, 黄恩华4, 吴昊5, 张玉海6(), 胡宗玉1()   

  1. 1.江苏中烟工业有限责任公司,江苏 南京 210004
    2.云南烟叶复烤有限责任公司楚雄复烤厂,云南 楚雄 675299
    3.华环国际烟草有限公司涡阳复烤厂,安徽 亳州 233606
    4.福建省三明金叶复烤有限责任公司,福建 三明 365001
    5.贵州烟叶复烤有限责任公司遵义复烤厂,贵州 遵义 563099
    6.郑州烟草研究院 烟草工艺研究室,河南 郑州 450000
  • 收稿日期:2025-01-09 出版日期:2026-03-11 发布日期:2026-03-30
  • 通讯作者: 张玉海,胡宗玉
  • 作者简介:胡宗玉,E-mail:huzy@jszygs.com
    张玉海,E-mail:yancaozhangyuhai@126.com
    许强,主要从事烟草加工工艺研究。E-mail:15077877253@163.com
  • 基金资助:
    江苏中烟工业有限责任公司科技项目(H202210);中国烟草总公司重点研发项目(110202102005);中国烟草总公司重点研发项目(11020222008);中国烟草总公司“揭榜挂帅”项目(110202103006)

The classification method of physical characteristics of tobacco leaves based on K-means clustering

XU Qiang1(), SU Xuemiao2, MIAO Chenlin1, ZHANG Yifan1, YANG Yue1, CHEN Chao3, HUANG Enhua4, WU Hao5, ZHANG Yuhai6(), HU Zongyu1()   

  1. 1.Jiangsu Tobacco Industry Co. ,Ltd. ,Nanjing 210004,Jiangsu
    2.Chuxiong Redrying Factory of Yunnan Tobacco Redrying Co. ,Ltd. ,Chuxiong 675299,Yunnan
    3.Guoyang Redrying Factory of Huahuan International Tobacco Co. ,Ltd. ,Bozhou 233606,Anhui
    4.Fujian Sanming Jinye Redrying Co. ,Ltd. ,Sanming 365001,Fujian
    5.Zunyi Redrying Factory of Guizhou Tobacco Redrying Co. ,Ltd. ,Zunyi 563099,Guizhou
    6.Tobacco Technology Research Laboratory,Zhengzhou Tobacco Research Institute,Zhengzhou 450000,Henan
  • Received:2025-01-09 Online:2026-03-11 Published:2026-03-30
  • Contact: ZHANG Yuhai,HU Zongyu

摘要:

为明确不同加工属性类型烟叶的典型产地、物理特性指标及其数据分布区间,探究不同加工属性类型烟叶的物理特性,本研究选取来自云南省、河南省等18个省份72个地级市31个品种的烟叶样品,分析了其黏附力、剪切强度、穿透强度、叶梗结合力、支脉结合力、拉力和伸长率等物理特性指标,采用K-means聚类和高斯混合模型算法进行分析,并建立加工属性分型方法。结果表明,采用正常系数评价指标可提高模型聚类效果,采用K-means模型的烟叶物理特性聚类效果要优于高斯混合模型,当以0.35作为正常系数的阈值时,可将烟叶样品物理指标按典型产地分为3类。本研究为确定不同加工属性类型烟叶的物理特性提供了一定的参考依据。

关键词: 烟叶, 物理特性, 聚类模型, K-means聚类模型, 高斯混合模型

Abstract:

To clarify the typical production areas,physical characteristic indicators and data distribution intervals of tobacco leaves with different processing attribute types,and to explore the physical characteristics of tobacco leaves with different processing attribute types,this study selected 31 varieties of tobacco leaves from 72 prefecture-level cities in 18 provinces including Yunnan Province and Henan Province. The physical characteristic indicators such as adhesion force,shear strength,penetration strength,leaf stem binding force,branch binding force,tensile force and elongation were analyzed. The K-means clustering and Gaussian mixture model algorithms were adopted for analysis,and the classification method of processing attributes was established. The results showed that using the normal coefficient evaluation index can improve the clustering effect of the model. The clustering effect of the physical characteristics of tobacco leaves using the K-means model was superior to that of the Gaussian mixture model. When 0.35 was taken as the threshold of the normal coefficient,the physical indicators of tobacco leaf samples can be classified into three categories according to typical production areas. This study provides a certain reference basis for determining the physical properties of tobacco leaves with different processing attribute types.

Key words: tobacco leaf, physical characteristic, clustering model, K-means clustering model, Gaussian mixture model

中图分类号: