中国农业科学 ›› 2019, Vol. 52 ›› Issue (9): 1518-1528.doi: 10.3864/j.issn.0578-1752.2019.09.004

• 耕作栽培·生理生化·农业信息技术 • 上一篇    下一篇

倒伏胁迫下玉米抽雄期叶面积密度光谱诊断

周龙飞1,2,3,4,顾晓鹤2,3,4(),成枢1,杨贵军2,3,4,孙乾1,2,3,4,束美艳1,2,3,4   

  1. 1 山东科技大学测绘科学与工程学院,山东青岛 266590
    2 农业部农业遥感机理与定量遥感重点实验室/北京农业信息技术研究中心,北京 100097
    3 国家农业信息化工程技术研究中心,北京 100097
    4 北京市农业物联网工程技术研究中心,北京 100097
  • 收稿日期:2018-12-11 接受日期:2019-01-30 出版日期:2019-05-01 发布日期:2019-05-16
  • 通讯作者: 顾晓鹤
  • 作者简介:周龙飞,E-mail: ZLF9510@163.com
  • 基金资助:
    国家自然科学基金(41571323);北京市自然科学基金(6172011);院创新能力建设专项(KJCX20170705)

Spectral Diagnosis of Leaf Area Density of Maize at Heading Stage Under Lodging Stress

ZHOU LongFei1,2,3,4,GU XiaoHe2,3,4(),CHENG Shu1,YANG GuiJun2,3,4,SUN Qian1,2,3,4,SHU MeiYan1,2,3,4   

  1. 1 College of Geomatics, Shandong University of Science and Technology, Qingdao 266590, Shandong
    2 Key Laboratory of Quantitative Remote Sensing in Agriculture of Ministry of Agriculture/Beijing Research Center for Information Technology in Agriculture, Beijing 100097
    3 National Engineering Research Center for Information Technology in Agriculture, Beijing 100097
    4 Beijing Engineering Research Center for Agriculture Internet of Things, Beijing 100097
  • Received:2018-12-11 Accepted:2019-01-30 Online:2019-05-01 Published:2019-05-16
  • Contact: XiaoHe GU

摘要:

【目的】叶面积密度(leaf area density,LAD)反映作物在垂直方向上体积内叶面积总量的差异,体现作物冠层内叶面积随着高度变化的分布状况。本文旨在探索玉米叶面积密度对于倒伏胁迫强度的表征能力及其光谱响应规律。【方法】以抽雄期倒伏夏玉米为研究对象,获取倒伏后玉米多期LAD及冠层光谱数据,对倒伏玉米冠层光谱进行一阶微分和小波变换处理,根据LAD与冠层光谱一阶微分及小波分解系数的相关性分析,筛选LAD敏感波段和最佳小波分解尺度,采用偏最小二乘法构建倒伏玉米LAD光谱诊断模型,并利用实测样本验证模型精度。【结果】玉米LAD随着倒伏胁迫程度的增强而增大,LAD可有效表征玉米倒伏胁迫强度及其自身恢复能力;玉米倒伏后冠层结构发生较大变化,倒伏玉米冠层光谱反射率较正常玉米整体增高,近红外波段的增幅相比可见光波段更高,倒伏强度越强则光谱反射率越高;LAD敏感波段主要分布在蓝光波段354—442、472—495 nm和红光波段649—829 nm以及近红外波段903—1 195 nm和1 564—1 581nm;同一阶微分处理相比,基于连续小波变换的玉米倒伏LAD诊断模型的验证R2提高6.08%—9.11%,RMSE降低23.08%—31.63%;小波分解尺度对LAD诊断精度有一定的影响,中低尺度模型精度优于高尺度模型,其中第5尺度构建的模型对LAD的拟合效果最优(R2=0.898,RMSE=1.016)。【结论】利用连续小波变换技术对玉米冠层高光谱解析,可有效诊断倒伏胁迫下的玉米叶面积密度,可以为玉米倒伏胁迫灾情遥感监测提供必要的先验知识。

关键词: 抽雄期, 倒伏胁迫, 连续小波变换, LAD, 玉米, 高光谱

Abstract:

【Objective】 Leaf area density (LAD) reflects the difference of the total leaf area per volume in vertical direction and the distribution of the leaf area in the canopy with the change of height. The purpose of this study was to explore the characterization ability of maize leaf area density and its spectral response to lodging stress intensity. 【Method】 Taking lodging summer maize at heading stage as the research object, the multi-stage of LAD and canopy spectral data after lodging were obtained. The first-order differential and wavelet transform of the canopy spectrum of lodging maize were processed. Based on the correlation analysis between LAD, the first-order differential and wavelet decomposition coefficients of canopy spectrum, the sensitive bands of LAD and the optimal wavelet decomposition scale were screened. Partial least squares (PLS) method was used to construct the LAD spectral diagnosis model of lodging maize, and the accuracy of the model was verified by the measured samples.【Result】The LAD of maize increased with the increase of lodging stress, and LAD could effectively characterize the intensity of lodging stress and recovery ability of maize. After lodging, the canopy structure of maize changed greatly. The spectral reflectance of lodging maize canopy was higher than that of normal maize. The increase of near infrared band was higher than that of visible band. The stronger lodging intensity was, the higher spectral reflectance was. The sensitive bands of LAD were mainly distributed in the blue band 354-442 nm and 472-495 nm, the red band 649-829 nm, and the near infrared band 903-1 195 nm and 1 564-1 581 nm. Comparing with the first-order differential, the validation R 2 of LAD diagnostic model of maize lodging based on continuous wavelet transform increased by 6.08%-9.11%, and RMSE decreased by 23.08%-31.63%. The scale of wavelet decomposition had a certain influence on the diagnostic accuracy of LAD. The accuracy of the low- and medium-scale model was better than that of the high-scale model, and the model constructed by the fifth scale had the best fitting effect on LAD (R 2=0.898, RMSE=1.016). 【Conclusion】 The application of continuous wavelet transform to analyze the maize canopy hyperspectral could effectively diagnose maize leaf area density under lodging stress. It could provide necessary prior knowledge for remote sensing monitoring of maize lodging stress disaster.

Key words: heading stage, lodging stress, continuous wavelet transform, LAD, maize, hyperspectral