摘要
叶片叶绿素含量是评价作物生长状况的重要指标。为实现玉米叶片叶绿素含量的准确、高效高光谱估测,以玉米大田试验为基础,于7月1日(大喇叭口期)、7月19日(灌浆初期)和8月18日(腊熟期)利用ASD高光谱仪和便携式叶绿素仪(SPAD-502)分别测定了玉米叶片高光谱数据和叶绿素含量相对值SPAD;利用连续投影算法提取出玉米叶片光谱的特征波长,再用BP神经网络构建SPAD值的估算模型,并对模型进行验证。结果表明,3个日期的分段监测模型及统一监测模型的R2分别为0.885,0.900,0.675,0.827;RMSE分别为2.156,2.103,3.236,2.651;7月1日模型、7月19日模型和统一监测模型均具有较高的精度,同时检验模型RPD均大于2,具有很好的预测能力;而8月18日的监测模型表现较差(RPD=1.641),但也达到可用水平。表明利用连续投影算法结合BP神经网络可以进行玉米叶片SPAD值的高光谱估算。
Leaf chlorophyll content is an important index to evaluate crop growth status. To achieve the estimation of maize leaves chlorophyll content accurately and efficiently through hyperspectral remote sensing, in this paper, based on the field experiments of maize, the hyperspectral data and the relative value of chlorophyll content of maize leaves were measured, respectively using ASD spectrometer and portable chlorophyll meter(SPAD-502) on July 1(large bell stage), July 19(early filling stage) and August 18(ripening stage). The characteristic wavelengths of the spectrum of maize leaves were extracted by successive projection algorithm, and the estimation models of SPAD value were constructed by BP neural network, and the models were verified. The results showed that R2 of the segmented monitoring models and the unified monitoring model of the three dates were 0.885, 0.900, 0.675 and 0.827, respectively, RMSE were 2.156, 2.103, 3.236 and 2.651, respectively. The models on July 1 , July 19 and the unified monitoring model all had high accuracy, and the RPD of the test models were greater than 2, indicating that the modles had good prediction ability. The monitoring model on August 18 performed poorly(RPD=1.641), but also reached the usable level. It shows that continuous projection algorithm combined with BP neural network can be used for estimation of hyperspectral estimation of SPAD value of maize leaves.
作者
董哲
杨武德
朱洪芬
郭琪
王子轩
白建伟
DONG Zhe;YANG Wude;ZHU Hongfen;GUO Qi;WANG Zixuan;BAI Jianwei(Institute of Dry Farming Engineering,Shanxi Agricultural University,Taigu 030801,China)
出处
《山西农业科学》
2019年第5期751-755,共5页
Journal of Shanxi Agricultural Sciences
基金
国家自然科学基金项目(31871571)
关键词
高光谱
玉米叶片
SPAD值
连续投影算法
BP神经网络
hypersprctral spectrum
maize leaves
SPAD value
continuous projection algorithm
BP neural network
