叶面施硒下荞麦冠层叶片氮含量遥感估测研究  

Remote Sensing Estimation of Buckwheat Leaf Nitrogen Content under Foliar Selenium Application

作　　者：马纬 武志明[1] 余科松 马晓光 MA Wei;WU Zhiming;YU Kesong;MA Xiaoguang(College of Agricultural Engineering,Shanxi Agricultural University,Jinzhong Shanxi 030801,China)

机构地区：[1]山西农业大学农业工程学院,山西晋中030801

出　　处：《农业工程》2023年第5期34-38,共5页AGRICULTURAL ENGINEERING

基　　金：省部共建有机旱作农业国家重点实验室(筹)自主研发课题(202105D121008-3-4)。

摘　　要：氮是植物生长发育过程中必需的营养元素之一,快速准确地获取其含量对大田农作物监测和管理有重要意义。采用无人机(UAV)搭载多光谱传感器对田间荞麦冠层叶片氮含量(Leaf Nitrogen Content,LNC)进行定量化估测,为荞麦叶片的信息化管理提供理论依据。试验选用不同荞麦品种为研究对象,通过无人机于荞麦开花期和灌浆期获取多光谱影像并同步采集荞麦冠层叶片的LNC,分别提取了5个波段下的反射率,选用与叶片LNC相关的12个植被指数进行皮尔逊(Pearson)相关性分析,选取17个光谱变量中相关性较高的特征变量与实测LNC进行PLSR、SVM和BPNN回归建模。结果表明,适量施用叶面硒肥可促进叶片吸收氮素从而增加LNC,过量硒肥不能持续提高LNC。G、R、NIR、NDVI、RDVI、RVI、SAVI、NLI、OSAVI、GRVI和LNC相关性较高,最高为GRVI,达到了0.824。采用BP神经网络建立的回归模型表现最优,盛花-灌浆期预测集决定系数(R^(2))为0.828,均方根误差(RMSE)为2.172,验证集R^(2)为0.939,RMSE为1.100,RPD为4.587。因此,无人机多光谱遥感技术可实现大田尺度的荞麦冠层叶片LNC估测。Nitrogen is one of a large number of elements in process of plant growth and development.The rapid and accurate acquisition of nitrogen content is of great significance for crop monitoring and management in the field.Leaf Nitrogen Content(LNC)of buckwheat canopy was quantified by UAV equipped with multispectral sensors.It could provide a theoretical basis for information management of buckwheat leaves.Taking different buckwheat varieties as research object,by using UAV to obtain multispectral images of buckwheat canopy leaves during flowering and filling stages,and synchronously collect LNC of buckwheat canopy leaves,reflectivity was extracted at 5 bands.12 vegetation indices related to LNC were selected for Pearson correlation analysis.Characteristic variables with higher correlation among 17 spectral variables were selected for PLSR,SVM,and BPNN regression modeling with measured LNC.G,R,NIR,NDVI,RDVI,RVI,SAVI,NLI,OSAVI,GRVI have a high correlation with LNC,with the highest being GRVI,reaching 0.824.Regression model established by BP neural network showed the best performance.Coefficient of determination(R^(2))and root mean square error(RMSE)of prediction set were 0.828 and 2.172,respectively.The R^(2),RMSE and RPD of validation set were 0.939,1.100 and 4.587 respectively.Therefore,the UAV multi-spectral remote sensing technology could realize field-scale LNC estimation of buckwheat canopy leaves.

关 键 词：荞麦 氮 遥感 无人机 多光谱传感器 土壤像元 

分 类 号：S127[农业科学—农业基础科学]