基于PCA-WNN的玉米叶片叶绿素含量遥感反演建模方法

The Remote Sensing Modeling Method on Inversing Chlorophyll Content of Maize Leaves Based on PCA-WNN

为寻找一种准确、非破坏性的叶绿素含量获取方法,实时掌握作物的生理状况,研究一种基于PCAWNN的玉米叶片叶绿素含量遥感反演模型。利用SVC HR-1024I光谱仪采集盆栽玉米叶片光谱,同时用SPAD-502便携式叶绿素计测定叶绿素含量。从包络线去除、微分处理后的光谱曲线中提取7个光谱特征参数(SCPs)并与修改型土壤调节植被指数(MSAVI)、归一化差值植被指数(NDVI)、修正植被指数(MVI)、比值植被指数(RVI)、差值植被指数(DVI)5种植被指数分别结合主成分分析(PCA),并提取前4个主分量作为小波神经网络(WNN)的输入因子,以Morlet母小波基函数作为激励函数,建立隐含层节点数为3的PCAWNN模型反演玉米叶片叶绿素含量。通过精度检验,表明7个SCPs与MSAVI组合的建模精度最高,验证小波神经网络反演玉米叶绿素含量的可行性以及其预测精度比BP神经网络更好。

In order to find an accurate and non destructive method of obtaining the chlorophyll content and understand well the real-time crop physiological state, a remote sensing model was presented to inverse the chlorophyll content of maize leaves based on PCA-WNN. The spectral reflectance of potted maize leaves was measured by SVC HR-1024I spectrometer and chlorophyll content were measured by SPAD-502 chlorophyll meter at the same time. Seven spectral characteristic parameters（SCPs） were extracted from the continuum removal and derivative spectra data, and combined with one of the MSAVI（modified soil adjusted vegetation index）, NDVI （normalized difference vegetation index）, MVI（modified vegetation index）,RVI（ratio vegetation index） and DVl（difference vegetation index） respectively. Then the principal component analysis（PCA） was used to extract the first four principal components as input factors of wavelet neural network（WNN）, the Morlet mother wavelet basis function was selected as the excitation function. Finally, the inverse model of PCA WNN was established with three hidden layer nodes to predict the chlorophyll content of maize leaves Through precision testing, the results showed that the inverse precision of the combination of the SCPs with MSAVI was the highest. It is verified feasible that using the WNN to inverse the chlorophyll content and its inverse precision was higher than BP neural network.