由粒子吸收光谱提取浮游植物吸收光谱的人工神经网络方法

基于现场实验数据集及人工神经网络技术,论文提出了一种从海中粒子吸收光谱提取浮游植物吸收光谱的方法。这个数据集包含了海中粒子吸收光谱和对应的浮游植物吸收光谱,并被分为三个子集:训练集、印证集和试验集。本研究所利用的人工神经网络系统为多层感知器,训练后的人工神经网络的性能由印证集和试验集来评价。实验结果表明,文中所提出的方法可成功地提取浮游植物的吸收光谱,其提取精度与传统的实验方法相当。

An Artificial Neural Network Method for Extraction of Phytoplankton Absorption Spectra from Total Particulate Absorption Spectra

In this paper,a method for extraction of phytoplankton absorption spectra from total particulate absorption spectra was proposed.It is derived from a database which contains the in situ measurements of total particulate absorption spectra and concomitant absorption spectra of phytoplankton determined chemically,and subsequent application of artificial neural network(ANN).The database was divided three subsets: training data,validation data(dependent on training data in geographical positions and cruises),and test data(independent on training data).The ANN used in this study is the so-called BP network with three layers.The performance of the trained ANN is assessed by applying it to the validation data and test data.The experiment results show this method is successful for estimation of phytoplankton absorption spectra,and its accuracy is comparable to the fiber glass filter method.