基于光谱变换的冬小麦冠层SPAD值估算模型研究

The Estimation Model of Winter Wheat Canopy SPAD Value Based on Spectral Transformation

建立精准、无损的冬小麦叶绿素含量监测模型,为冬小麦施肥调控和产量评估提供科学依据。本研究以实测的冬小麦冠层光谱反射率和冠层SPAD值数据为基础,分析经变换后反射率与冠层SPAD值的相关关系,以此为基础构建了基于不同光谱变换形式下冬小麦全生育时段的冠层SPAD值监测模型。结果表明:以光谱变换后反射率与监测冠层SPAD值相关性最高为原则,筛选出的适宜光谱变换形式下的敏感波段,其中拔节-抽穗期为一阶微分(503 nm),抽穗-灌浆期为一阶微分(543 nm),灌浆-成熟期为对数的一阶微分(726 nm),全生育期为除以R930(724 nm);考虑各生育时段特点构建不同光谱变换形式下适宜的回归模型组合即模型组合1,考虑模型简单实用性构建单个光谱变换值的一元二次回归模型即模型2,模型1在各生育期及全生育期的决定系数R^2分别为0.836、0.855、0.917、0.890,且较同期模型2的决定系数R^2分别提高了24.4%、6.1%、57.8%、37.8%,表明采用各生育期不同光谱变换形式下适宜模型组合的监测效果优于单个光谱变换值的一元二次回归模型。

A precise and non-destructive monitoring model of chlorophyll content in winter wheat is established to provide scientific basis for the fertilization regulation and yield evaluation of winter wheat.Based on the measured spectral reflectance and SPAD value of winter wheat canopy,the correlation between the spectral reflectance and SPAD value of winter wheat canopy after transformation is analyzed.Based on this,a monitoring model of SPAD value of winter wheat canopy during the whole growth period under different spectral transformation forms is constructed.The results show that the sensitive bands in the form of spectral transformation are selected according to the principle of the highest correlation between reflectance after spectral transformation and SPAD value of monitoring canopy.Among them,jointing-heading stage is a first-order differential(503 nm),heading-filling stage is a first-order differential(543 nm),and filling-maturing stage is a first-order differential(726 nm).The whole growth period is divided by R930(724 nm).Considering the characteristics of each growth period,a suitable regression model combination under different spectral transformation forms is constructed,i.e.model combination 1.Considering the simplicity and practicability of the model,a one-dimensional quadratic regression model of single spectral transformation value,i.e.model 2,is constructed.The determinant coefficients R^2 of model 1 in each growth period and the whole growth period are 0.836,0.855,0.917 and 0.890,respectively.Compared with model 2 of the same period,the determinant coefficients R^2 increases by 24.4%,6.1%,57.8%and 37.8%respectively,which indicates that the monitoring effect of suitable model combination under different spectral transformation forms in different growth stages is better than that of single spectral transformation value quadratic regression model.