摘要
【目的】构建叶色参数与鲜烟叶片成熟度的关联模型,为精准判断大田烟草鲜叶采收成熟度提供技术支撑。【方法】运用多元回归及BP神经网络构建叶色参数-烟叶成熟度判定模型,通过比较不同参数体系和不同建模方式对不同叶位鲜烟叶片成熟度的判定准确度,筛选最优成熟度智能化判定模型。【结果】以叶色偏态分布复合参数集(N2,共33个参数)作为输入因子,基于BP神经网络构建的F4(33-10-1)模型是最佳的叶片成熟度判定模型,可以满足对不同叶位不同成熟度的判定精度需求,对下部、中部和上部烟叶成熟度的判定精度分别为84.44%、96.10%和92.56%,总体精度达94.15%。【结论】采用叶色偏态分布复合参数集(N2)作为输入因子且采用BP神经网络构建的不同叶位鲜烟叶成熟度的判定方法可在一定程度上解决烟草叶片成熟度田间判定准确度低的问题,且可为烟叶生产过程中鲜烟成熟度的智能化识别提供新思路。
【Objective】The correlation model between leaf color parameters and fresh tobacco leaf maturity is established to provide the technical support for accurately judging harvest maturity of tobacco leaves in tobacco field.【Method】The judgment model of leaf color parameters-tobacco leaf maturity is established by the multiple regression method and BP neural network.The optimal intelligent judgment model of tobacco leaf maturity is screened by comparing the maturity judgment accuracy of fresh tobacco leaves from different leaf position determined by different parameter systems and different modelling methods.【Result】The established F_(4)(33-10-1)model based on BP neural network and skewed distribution complex parameter set of leaf color(N_(2))is the optimal judgment model of tobacco leaf maturity,which can meet judgment accuracy requirement of different maturity of tobacco leaves from different leaf position.The maturity judgment accuracy of lower,middle and upper leaves determined by F_(4) model reaches 84.44%,96.10%and 92.56%respectively and the overall accuracy is up to 94.15%.【Conclusion】The established F_(4)(33-10-1)model based on BP neural network and skewed distribution complex parameter set of leaf color(N_(2))can solve the lower judgment accuracy of tobacco leaf maturity in tobacco field to some extent and provide a new idea for intelligent identification of fresh tobacco leaf maturity in actual tobacco production.
作者
林天然
姚争毅
常鹏飞
郑婧
李满玲
谢萍
陈诗瑶
储伟杰
李仁忠
LIN Tianran;YAO Zhengyi;CHANG Pengfei;ZHENG Jing;LI Manling;XIE Ping;CHEN Shiyao;CHU Weijie;LI Renzhong(Longyan Branch,Fujian Tobacco Company,Longyan,Fujian 364000;Fujian Tobacco Company,China Tobacco Corporation,Fuzhou,Fujian 350003;Nanjing Ecological Agriculture and Forestry Development(Southern Nanjing)Co.,Ltd.,Nanjing,Jiangsu 211500;Sanming Meteorological Bureau,Sanming,Fujian 365000,China)
出处
《贵州农业科学》
CAS
2022年第8期134-141,共8页
Guizhou Agricultural Sciences
基金
中国烟草总公司福建省公司科技项目“基于大数据的龙岩烟草精准农业研究与应用”(2021350000240020)
福建省烟草公司龙岩市公司科技项目“基于无人机平台的烟叶智能识别及云烟87成熟度判定系统的研发”(2020Y01)。
关键词
烟草鲜叶
成熟度
图像分割
BP神经网络
偏态分布
fresh tobacco leaf
maturity
image segmentation
BP neutral network
skewed distribution
