干旱胁迫下旺长期烤烟冠层叶绿素密度的高光谱估测

Hyperspectral Estimation of Canopy Chlorophyll Density in Flue-cured Tobacco under Different Drought Stress at the Vigorous Growth Stage

为精准、实时、无损地估算烤烟冠层叶绿素密度,快速获取烤烟光合性能与营养状况,基于不同程度干旱胁迫处理,采用ASD光谱仪,在综合分析群体原始高光谱反射率、一阶导数光谱反射率及已有光谱指数与冠层叶绿素密度(CCD)关系的基础上,建立烤烟CCD估算模型。结果表明:(1)干旱胁迫后烤烟冠层光谱反射率随叶绿素密度呈现规律性变化。(2)712 nm处的一阶导数与CCD相关性最好(r=0.838)。(3)利用一阶导数光谱建立的反演叶绿素密度的线性模型和BP神经网络模型中,均以BP神经网络模型效果最好,其模型决定系数R^2为0.9686,均方根误差RMSE 0.0778,表明模型的精度和稳定性均较好。研究结果可为实时监测旺长期烤烟群体光合能力及水分胁迫状况提供栽培管理依据。

This study aims to estimate the canopy chlorophyll density(CCD)of flue-cured tobacco with non destructive and accuratemethods in real time to obtain the photosynthetic capacity and nutritional status.To this end,the relationship of canopy chlorophylldensity and hyperspectral reflectance of flue-cured tobacco under different drought stresses was studied using an ASD spectrometer.Estimating models of the flue-cured tobacco canopy chlorophyll density were set up by means of the first derivative spectral reflectance.The results indicated that flue-cured tobacco canopy spectral reflectance showed orderly changes with chlorophyll density after droughtstresses.The correlation between the first derivative spectral reflectance at712nm and chlorophyll density is the best(r=0.838).TheBP neural network generated the best estimation.In the inversion of monadic linear model and BP neural network model for CCDusing the first derivative spectral reflectance,the BP neural network model showed the best effect with the R2reached0.9686andRMSE being0.0778.The results may provide the basis for the cultivation and management through long-term real-time monitoring ofthe photosynthetic capacity and water stress status of the flue-cured tobacco flourishing population.