基于主成分分析和BP神经网络的法国梧桐叶绿素含量高光谱反演研究

Research on platanus orientalis L.chlorophyll concentration estimation with hyperspectral data based on BP-artificial neural network and principal component analysis

利用ASD便携式野外光谱仪和SPAD-502叶绿素计实测了落叶阔叶树法国梧桐叶片的高光谱反射率与叶片绿度,并对原始光谱反射率及一阶导数光谱与叶片绿度进行了相关分析;综合分析了10个常见光谱植被指数与法国梧桐叶绿素含量的相关性与预测性;最后利用主成分分析对光谱数据进行降维,将得到的主成分得分作为BP人工神经网络模型的输入变量进行了法国梧桐叶绿素含量的估算。结果表明:法国梧桐的叶片反射光谱数据与叶绿素含量的相关性在可见光区域显著,导数光谱数据在绿黄光区和红光区的部分波段与叶绿素含量的相关系数大于对应波段光谱反射率与叶绿素含量的相关关系。在所列举的10个常用植被指数中归一化植被指数与叶绿素含量的关系最密切,相关系数达到了0.7957。主成分分析的BP神经网络模型可以容纳更多的波段信息进行叶绿素含量的估算,预测值与实测值之间的线性回归的确定性系数R2为0.9883,是一种良好的植被叶绿素含量高光谱反演模式。

Hyperspectral reflectance and green degree of Platanus orientalis L. ＇leaves were measured by the ASD portable spectrometer and the portable chlorophyll meter SPAD-502, respectively, and the relativity between hyperspectral reflectance, first derivative spectral reflectance and leaf green degree were analyzed. In order to estimate chlorophyll content and establish the best estimation model, ten normal vegetation indexes （VI） were compared to find out the best one. Then the principal component analysis （PCA） was used to reduce the dimensions of data, while maintaining the data characteristic effectively, and the principal component scores were used as the input variable of ANN-BP to estimate the content of chlorophyll. The results show that Platanus orientalis L. hyperspectral reflectance has a notable relationship with chlorophyll concentration in visible region, while the relationship between the first derivative spectral reflectance and chlorophyll concentration is better than the relationship of the hyperspectral reflectance with chlorophyll concentration in the part of green-yellow and red region. NDVI has the closest relation to chlorophyll concentration of all the ten normal VI in this article, and the correlation coefficient is 0. 7957. The ANN-BP based on PCA can contain more band information to estimate chlorophyll content, and the determination coefficient R2 between the predicted and the measured Platanus orientalis L ＇ chlorophyll content were 0. 9883, so the ANN-BP based on PCA in this article is a good method to be applied to hyperspectral data for estimation of Platanus orientalis L. chlorophyll concentration.