欢迎访问《中国农学通报》，
中国农学通报

中国农学通报 ›› 2024, Vol. 40 ›› Issue (1): 151-156.doi: 10.11924/j.issn.1000-6850.casb2023-0031

• 农业信息·科技教育 • 上一篇    下一篇

基于水分光谱指数的烟草叶片等效水厚度估测

贾方方1,2(), 滕世华3(), 何琳1, 付安旗1, 陈淑萍1, 赵中原1   

  1. 1 商丘师范学院生物与食品学院，河南商丘 476000
    2 郑州大学信息工程学院，郑州 450001
    3 云南香料烟有限责任公司，云南保山 678000
  • 收稿日期:2023-01-10 修回日期:2023-08-30 出版日期:2024-01-05 发布日期:2023-12-29
  • 通讯作者:
    滕世华，男，1982年出生，山东莘县人，助理农艺师，本科，主要从事烟叶生产技术推广工作。通信地址：678000 云南省保山市隆阳区建设路174号，Tel：0875-2166716，E-mail：。
  • 作者简介:

    贾方方，女，1984年出生，河南新郑人，副教授，博士，主要从事农业信息技术研究。通信地址：476000 河南省商丘市梁园区文化路298号，Tel：0370-3115990，E-mail：。

  • 基金资助:
    河南省烟草公司平顶山市公司科技项目“烟草黑胫病拮抗菌的筛选、鉴定及生物防治研究”(PYKJ202101); 河南省科技攻关项目“近红外高光谱成像对辣椒损伤的快速判别研究”(232102110280)

Estimating Equivalent Water Thickness of Tobacco Leaves Based on Water Hyperspectral Indices

JIA Fangfang1,2(), TENG Shihua3(), HE Lin1, FU Anqi1, CHEN Shuping1, ZHAO Zhongyuan1   

  1. 1 Department of Biology and Food, Shangqiu Normal University, Shangqiu, Henan 476000
    2 School of Information Engineering, Zhengzhou University, Zhengzhou 450001
    3 Yunnan Oriental Tobacco Co., Ltd., Baoshan, Yunnan 678000
  • Received:2023-01-10 Revised:2023-08-30 Published-:2024-01-05 Online:2023-12-29

RichHTML

2

PDF (PC)

33

摘要：

为及时、准确监测烟草叶片的水分状况，连续2年开展不同基因型烤烟品种的水分胁迫试验，测定烟叶的光谱反射率和叶片等效水厚度(EWT)，系统分析350~2500 nm波段范围内任意2个波段组合而成的比值水分指数(SRWI)和归一化水分指数(NDWI)，并构建烟草EWT的光谱指数预测模型。结果表明：（1）不同基因型烟草的叶片等效水厚度均随灌水量的减少而降低。（2）不同水分处理的烟叶光谱反射率在可见光波段和近红外波段均发生了规律性变化。（3）筛选出的烟草叶片等效水厚度的光谱敏感波段主要集中在可见光区域的500~600 nm、近红外区域的700~900和1000~1250 nm、短波红外区域的1900~2000 nm，最佳水分光谱指数分别为NDWI(R1920，R1930)、SRWI(R1930，R1920)，核心波段为1920、1930 nm。（4）利用水分光谱指数构建的烟草叶片等效水厚度的线性和非线性预测模型，以极限学习机模型(ELM)的精准度和稳定性最佳(P-R2=0.853**，T-R2=0.855**，RMSE=0.004)。表明可利用水分光谱指数NDWI(R1920，R1930)、SRWI(R1930，R1920)结合机器学习模型实现烟草叶片含水量的精准监测。

关键词: 烟草叶片, 等效水厚度, 水分光谱指数, 预测模型, 极限学习机

Abstract:

In order to timely and accurately monitor the tobacco leaf moisture status, this study conducted water stress experiments on different flue-cured tobacco varieties for over two consecutive years (2020 and 2021). The spectral reflectance and leaf equivalent water thickness (EWT) of tobacco leaves were measured, and the water spectral indices based on EWT were screened out and used to construct a prediction model. The results showed that: (1) the leaf equivalent water thickness of different tobacco genotypes decreased with the reduction of irrigation amount. (2) The spectral reflectance of tobacco leaves under different moisture treatments varied regularly in the visible and near-infrared wavelength ranges. (3) The spectral sensitive regions of EWT were mainly concentrated in the visible region of 500-600 nm, the near-infrared region of 700-900 nm and 1000-1250 nm, and the short-wave infrared region of 1900-2000 nm. The optimal water spectral indices were NDWI (R1920, R1930) and SRWI (R1930, R1920), and the core bands for EWT were 1920 nm and 1930 nm. (4) The accuracy and stability result of ELM was the best in the different linear and nonlinear prediction models of EWT prediction models, with the model decision coefficient of 0.853, the validation decision coefficient of 0.855, and the RMSE of 0.004. This indicates that water spectral indices combined with nonlinear models can be utilized to predict the moisture content in tobacco leaves.

Key words: tobacco leaf, equivalent water thickness, water hyperspectral characteristic parameters, prediction model, extreme learning machine