基于光谱变换定量估算苹果叶片的等效水厚度

运用高光谱技术快速无损地估算了苹果叶片的等效水厚度(Equivalent Water Thickness, EWT)，为苹果树的长势及干旱预警提供参考。以山东省烟台栖霞市红富士苹果树叶片为试验材料，在测定苹果叶片的光谱反射率和计算叶片EWT的基础上，分析了苹果叶片的EWT、原始光谱的反射率及其13种变换光谱反射率之间的相关性。筛选敏感波长后，建立了苹果叶片EWT的支持向量机定量的估算模型。13种光谱变换中，一阶导数(the First Derivative, FDR)、平方根的一阶导数(the First Derivative of the Square Root, FD(SqrtR))及倒数的对数的一阶导数(the First Derivative of the Logarithm of the Reciprocal, FDLg(1/R)])三种变换的相关性较好。确定了估测苹果叶片EWT的敏感波长。基于支持向量机回归分析方法，建立了定量估算叶片EWT的模型，验证集的决定系数R2达到了0.8147，相对分析误差(Relative Percent Deviation, RPD)达到了2.2671。结果表明，该模型具有较高的估测能力，支持向量机回归方法比较适于估算苹果叶片的EWT。该方法为利用高光谱技术定量估算苹果的生长状况提供了技术支撑。

Quantitative Estimation of Equivalent Water Thickness of Apple Leaves Based on Hyperspectral Transformation

A hyperspectral technology was utilized to estimate the Equivalent Water Thickness (EWT) of apple leaves in a rapid and nondestructive way, which could provide a reference for the early warning of growth and drought of apple trees. Taking the leaves of Red Fuji apple trees in Qixia, Yantai, Shandong Province as experimental materials, the correlation among EWT, original spectral reflectance and 13 kinds of transformation spectral reflectance was analyzed on the basis of the spectral reflectance of apple tree leaves and the calculation of their EWT. After the sensitive wavelengths were screened out, an estimation model of the EWT of apple tree leaves based on support vector regression was established. Among the 13 kinds of spectral transformation, 3 kinds of transformation such as the first derivative (FDR), the first derivative of the square root (FD(SqrtR)) and the first derivative of the logarithm of the reciprocal (FDLg(1/R)]) had better correlation. The sensitive wavelengths for estimating the EWT of apple tree leaves were determined. On the basis of the support vector regression, a mode for quantitatively estimating the blade EWT was established. The coefficient of determination R2 of the validation set reached 0.8147 and the relative percent deviation (RPD) reached 2.2671. The results showed that the model had better estimation ability and the support vector regression was more suitable to estimate the EWT of apple tree leaves. This method provided technical support for the quantitative estimation of the growth status of apples by hyperspectral techniques.