基于人工神经网络的海水漫射衰减系数的遥感反演方法

A Remote Sensing Method for the Determination of Seawater Diffuse Attenuation Coefficient Based on Artificial Neural Networks

利用NOMAD数据集建立了基于人工神经网络的漫射衰减系数Kd490的反演算法。该人工神经网络是3层的反向传输神经网络。其结构为输入层有4个节点,它们分别对应4个波段443,490,555,665 nm的遥感反射比,隐含层有10个节点,输出层1个节点对应于漫衰减系数Kd490。利用另一独立的现场测量数据集(COASTLOOC)印证该反演算法的性能。结果表明,该研究建立的反演算法的性能明显好于业务化SeaWiFS算法,略好于Lee等人的半分析算法。

An ANN-based algorithm for the retrieval of Ka490 was developed based on NOMAD database. The ANN employed in this study has three layers： an input layer with four neurons corresponding to the remote sensing reflectances Rrs443, Rrs490, Rrs555 and Rrs665, a hidden layer with ten neurons, and an output layer with one neuron corresponding to Kd490. Another independent in-situ data set （COASTLOOC） is used to validate the performance of the ANN-based algorithm. The results show that the performance of the algorithm developed in this study is much better than that of SeaWiFS operational algorithm, and slightly better than that of the algorithm developed by Lee et al. （2005）.