|
|
| |
|
|
码上扫一扫！ |
| 烟叶素质采集装置设计与试验 |
|
李昌根1, 吴飞跃2, 李静超3, 王廷贤3, 何健2, 林勇3, 魏硕1, 杨占伟1,3
|
|
1.河南农业大学 烟草学院，河南 郑州 450046;2.福建中烟工业有限责任公司，福建 厦门 361012;3.福建省烟草公司 南平市公司，福建 南平 353000
|
|
| 摘要: |
| 【目的】设计一种烟叶素质采集装置，为实现烟叶素质特征的快速准确采集提供支持。【方法】设计了主要由箱体、光源、CCD工业相机、标准色卡、质量传感器等组成的烟叶素质采集装置，并应用机器视觉技术搭建了烟叶素质识别系统。对该采集装置光源的照明控制值（illumination control value，ICV）进行优化后，使用该装置采集烟叶图像，提取并分析不同成熟度（欠熟、适熟和过熟）烟叶的颜色特征（RGB和HSV颜色空间各通道值的一阶矩（MeR、MeG、MeB、MeH、MeS、MeV）及二阶矩（VaR、VaG、VaB、VaH、VaS、VaV）），再分别运用决策树学习算法、随机森林算法以及BP神经网络建立成熟度识别模型，通过识别单片烟叶及单竿烟叶成熟度，对各算法识别模型的性能进行评估。【结果】烟叶素质采集装置光源的最佳照明控制值（ICV）为90。在此条件下，单片及单竿烟叶的VaR和VaS及单片烟叶的MeR、MeS在不同成熟度之间差异不显著；与欠熟烟叶相比，适熟与过熟单片烟叶的MeG、MeB、MeV及单竿烟叶的MeR、MeG、MeB和MeV均显著增大（P<0.05），单片烟叶的MeH、VaG、VaB、VaH和VaV及单竿烟叶的MeH、MeS、VaG、VaB、VaH和VaV均显著减小（P<0.05），其中单片和单竿过熟烟叶的MeG、MeB和MeV分别增大了6.06%，16.30%，6.28%和4.13%，13.34%，4.14%，MeH、VaG、VaB、VaH和VaV分别减少了15.95%，20.16%，33.63%，15.14%，21.18%和16.86%，17.81%，15.11%，7.90%，18.03%。所建立的烟叶成熟度识别模型中，决策树学习算法最优，其对单片烟叶和单竿烟叶识别的平均准确率分别为97%和95%。【结论】所设计的烟叶素质采集装置可以精准高效采集烟叶的素质特征，为烟叶素质的在线自动识别提供了一种有效工具。 |
| 关键词: 烟叶 成熟度识别 素质特征 图像采集 采集装置设计 |
| DOI：DOI:10.13207/j.cnki.jnwafu.2025.06.015 |
| 分类号: |
| 基金项目:中国烟草总公司科技重点研发项目（110202102007）；福建省烟草公司南平市公司资助项目（NYK2023-03-03） |
|
| Design and experiment on collection device for the quality of tobacco leaves |
|
LI Changgen1, WU Feiyue2, LI Jingchao3, WANG Tingxian3, HE Jian2, LIN Yong3, WEI Shuo1, YANG Zhanwei1,3
|
|
1.College of Tobacco Science,Henan Agricultural University，Zhengzhou,Henan 450046,China;2.China Tobacco Fujian Industry Co.,Ltd.Xiamen,Fujian 361012,China;3.Nanping Branch of Fujian Provincial Tobacco Company,Nanping,Fujian 353000,China
|
| Abstract: |
| 【Objective】The study aimed to design a tobacco leaf quality collection device to provide support for achieving rapid and accurate collection of tobacco leaf quality characteristics.【Method】A tobacco leaf quality collection device composed of a box,a light source,a CCD industrial camera,a standard color card,a quality sensor and other components,was developed.A tobacco quality recognition system was also built using machine vision technology.After optimizing the illumination control value (ICV) of the device,it was used to collect tobacco leaf images,extract and analyze the color characteristics of different maturity tobacco leaves (underripe,ripe and overripe) (the first moment (MeR,MeG,MeB,MeH,MeS,MeV) and second moment (VaR,VaG,VaB,VaH,VaS,VaV) of each channel value in RGB and HSV color space).Then the maturity recognition model was established by using the decision tree learning algorithm,random forest algorithm and BP neural network,respectively.The performance of each algorithm recognition model was evaluated by identifying the maturity of single tobacco leaves and single pole tobacco leaves.【Result】 The optimal illumination control value (ICV) of the light source of the tobacco leaf quality acquisition device was 90.Under this condition,there was no significant difference in VaR and VaS of single leaves and single pole tobacco leaves and MeR and MeS of single tobacco leaves at different stages of maturity.Compared with the underripe tobacco leaves,the MeG,MeB and MeV of the ripe and overripe single leaves and the MeR,MeG,MeB and MeV of the single tobacco leaves were significantly increased (P<0.05),while the MeH,VaG,VaB,VaH and VaV of the single tobacco leaves and the MeH,MeS,VaG,VaB,VaH and VaV of the single tobacco leaves were significantly decreased (P<0.05).Specifically,the MeG,MeB and MeV of single-leaf and single-pole overripe tobacco leaves increased by 6.06%,16.30%,6.28% and 4.13%,13.34%,4.14%,respectively.The MeH,VaG,VaB,VaH and VaV decreased by 15.95%,20.16%,33.63%,15.14%,21.18% and 16.86%,17.81%,15.11%,7.90%,18.03%,respectively.In the established recognition model,the decision tree learning algorithm was the best,and the average accuracy of single-leaf tobacco leaves and single-pole tobacco leaves recognition was 97% and 95%,respectively.【Conclusion】The designed tobacco leaf quality acquisition device can accurately and efficiently collect the quality characteristics of tobacco leaves,and provide an effective tool for online automatic identification of tobacco leaf quality. |
| Key words: tobacco leaves maturity identification quality features image acquisition design of collection device |
|
|
|
|
