基于无人机高光谱影像的玉米叶片氮素含量估算

Estimation of Leaf Nitrogen Content in Maize based on UAV Hyperspectral Image

利用无人机采集影像进行田块尺度的作物氮素含量估算,因其具有无破坏性、时效性强等优势而备受关注。东北黑土区作为我国重要的粮食主产区,精准获取作物的氮素含量对国家粮食安全具有重要意义。研究基于玉米拔节期、吐丝期和成熟期的无人机高光谱影像,提取22种窄波段光谱指数,通过逐步回归方法对黑土区玉米叶片的氮素含量(Leaf Nitrogen Content,LNC)进行定量估算。结果表明:本研究中3个生育期的玉米LNC估算模型均具有较高的估算精度,其中拔节期的估算精度最高,其决定系数(R^(2))、均方根误差(RMSE)、归一化均方根误差(nRMSE)分别为0.76、0.31%、0.15%;吐丝期的估算精度最低,相应的3个指标分别为0.33、0.27%、0.19%。通过逐步回归模型筛选出在各生育期指示作用较强的光谱指数,拔节期为VARI(Vegetation Atmospherically Resistant Index)、DDI(Desertification Difference Index)、EVI(Enhanced Vegetation Index);吐丝期为MTCI(MERIS Terrestrial Chlorophyll Index)、SIPI(Simple Insensitive Pigment Index);成熟期为EVI(Enhanced Vegetation Index)、CCI(Canopy Chlorophyll Index)、NDVI(Normalized Difference Vegetation Index)。最终,利用在玉米各生育期估算精度最高的模型获得玉米叶片氮素含量的空间分布图,其空间分布特征与玉米LNC实测情况相一致,而氮肥施用量对不同田间处理的小区间玉米LNC的影响较大。综上,本研究结果可为东北黑土区玉米叶片氮素含量的无损、快速、动态监测提供数据基础与决策支持。

The estimation of crop nitrogen content at the field scale by using Unmanned Aerial Vehicle(UAV)images has attracted increasing attention due to their nondestructiveness and time-effectiveness.The black soil region of Northeast China is the main agricultural production base in China,and accurately obtaining crop nitro⁃gen content is of great significance for national food security.In this study,the Leaf Nitrogen Content(LNC)of maize was estimated by the stepwise regression method using UAV hyperspectral images and 22 narrowband spectral indices at the jointing,silking,and maturity growth stages of maize.The results showed that the maize LNC estimation models at the three growth stages all had good performance.Moreover,the estimation accura⁃cy of the model at the maturity stage was slightly higher than those from the other two stages,with R^(2),RMSE,and nRMSE values of 0.76,0.31%,and 0.15%,respectively.The estimation model at the silking stage had the lowest accuracy,with R^(2),RMSE,and nRMSE values of 0.33,0.27%,and 0.19%,respectively.At the same time,the spectral indices that can indicate maize LNC were obtained.They were VARI(Vegetation At⁃mospherically Resistant Index),DDI(Desertification Difference Index)and EVI(Enhanced Vegetation In⁃dex)at the jointing stage;MTCI(MERIS Terrestrial Chlorophyll Index)and SIPI(Simple Insensitive Pig⁃ment Index)at the silking stage;and EVI(Enhanced Vegetation Index),CCI(Canopy Chlorophyll Index)and NDVI(Normalized Difference Vegetation Index)at the maturity stage.Finally,the spatial distribution map of maize LNC was obtained using the model with the highest estimation accuracy at each growth stage,and its spatial distribution characteristics were consistent with the actual maize LNC conditions.However,the amount of nitrogen fertilizer had a greater impact on the maize LNC among the microplots with different treat⁃ments.The results of this study can provide a database and decision support for the nondestructive,rapid and dy⁃namic monitoring of maize leaf nitrogen content in the black soil region of Northeast China.