基于PCA和BP的水体Chl-a含量高光谱反演方法

Method of Hyperspectral Inversion for Water Chl-a Content Based on PCA and BP

利用主成分分析法对高光谱数据进行降维,将主成分得分作为输入,将水体20个采样点的叶绿素a(Chl-a)含量实测数据作为输出,对BP神经网络进行训练学习,实现压缩光谱数据与Chl-a含量的自适应非线性映射,并利用另外10个采样点数据对网络进行验证,结果表明预测值与实测值差距较小。

Dimensionality of hyperspectrum data was reduced by using PCA.The principal component scores were used as input,and the measured data for Chl-a content of twenty water sampling points were used as output.The BP Neural Networks were trained and studied.Adaptive nonlinear mapping on compressed spectral data and Chl-a content was realized.It was verified by using the other 10 sampling points data on network,and the results show that the gap between forecast value and actual value is small.