基于高光谱遥感的小麦子粒谷氨酰胺合成酶活性估算研究  

Estimation of glutamine synthetase activity in wheat grain based on hyperspectral remote sensing

作　　者：熊淑萍[1] 丁绍强 郭建彪[1] 张志勇[1] 徐赛俊 樊泽华 穆彦玲 马新明[1] XIONG Shuping;DING Shaoqiang;GUO Jianbiao;ZHANG Zhiyong;XU Saijun;FAN Zehua;MU Yanling;MA Xinming(College of Agronomy,Henan Agricultural University,Zhengzhou 450002,China)

机构地区：[1]河南农业大学农学院,河南郑州450002

出　　处：《河南农业大学学报》2021年第5期821-829,共9页Journal of Henan Agricultural University

基　　金：河南省小麦产业技术体系项目(S2010-01-G04);河南省高等学校重点科研项目(21A210015);国家重点研发计划课题(2016YFD0300205)。

摘　　要：为利用高光谱遥感技术快速、无损、准确估算小麦子粒中谷氨酰胺合成酶(GS)活性,设置不同小麦品种和氮肥处理组合大田试验,以小麦花后10和20 d子粒中GS酶活性为研究对象,同时测定相应时期小麦冠层的高光谱特征,通过一阶导数、二阶导数和多元散射校正3种方法,对小麦冠层原始光谱进行预处理,分析原始光谱、一阶导数、二阶导数和多元散射校正与小麦子粒GS活性的相关性,并以此为输入,利用偏最小二乘回归、支持向量机回归和BP人工神经网络3种方法,构建了小麦子粒GS活性的高光谱遥感估算模型,运用决定系数(R^(2))和均方根误差(RMSE)对模型进行评价。结果表明,经微分(一阶导数、二阶导数)预处理后小麦冠层光谱与小麦子粒GS活性的相关性优于原始光谱和多元散射校正,其所构建的估算模型精度明显高于原始光谱和多元散射校正,尤以基于一阶导数光谱的偏最小二乘法估算模型表现最好,其模型建模集的R^(2)和RMSE分别为0.942,0.025 4,验证集的R^(2)和RMSE分别为0.755,0.034 0,具有良好的估算精度和应用潜力。In order to rapidly,nondestructively and accurately estimate the activity of glutamic synthase in wheat grains by using hyperspectral remote sensing technology,a field experiment was set with different wheat varieties and nitrogen fertilizer combinations.The GS activity in wheat grains on 10 d and 20 d after anthesis was taken as the research object,and the hyperspectral characteristics of wheat canopies of the corresponding period were measured at the same time.Three methods,first derivative,second derivative and multiple scattering correction,were used to pretreat the original spectra of wheat canopies.By analyzing the correlation between the original spectrum,the first derivative,the second derivative and the multiple scattering correction and the GS activity of wheat grains,a hyperspectral remote sensing estimation model of the GS activity of wheat grains was constructed by using three model building methods,the partial least squares regression,support vector machine regression and BP artificial neural network.The determination coefficient(R^(2))and root mean square error(RMSE)were used to evaluate the model.The results showed that the correlation between wheat canopy spectra and GS activity after differential(first derivative,second derivative)pretreatment was better than that of the original spectrum and multiple scattering correction,and the accuracy of the estimation model was significantly higher than that of the original spectrum and the multiple scattering correction.Especially,the partial least squares estimation model based on the first derivative spectrum performed the best.The R^(2) and RMSE of the modeling set was 0.942 and 0.0254,respectively,and the R^(2) and RMSE of the verification set was 0.755 and 0.0340,respectively,indicating that the model has good estimation accuracy and application potential.

关 键 词：小麦 谷氨酰胺合成酶活性 高光谱 预处理 模型 

分 类 号：S512.1[农业科学—作物学]