基于人工神经网络的一类水域叶绿素-a浓度反演方法

张亭禄; 贺明霞

doi:10.11834/jrs.20020108

浏览量 : 0 下载量: 965 CSCD: 14

PDF
导出
分享
收藏
专辑

基于人工神经网络的一类水域叶绿素-a浓度反演方法
A Method to Retrieve the Oceanic Chlorophyll-a Concentrations in Case I Water Based on Artificial Neural Network
2002年第1期页码：40-44
纸质出版：2002
DOI： 10.11834/jrs.20020108
稿件说明：

移动端阅览

[1]张亭禄,贺明霞.基于人工神经网络的一类水域叶绿素-a浓度反演方法[J].遥感学报,2002(01):40-44. DOI： 10.11834/jrs.20020108.

ZHANG Ting_lu, HE Ming_xia. A Method to Retrieve the Oceanic Chlorophyll-a Concentrations in Case I Water Based on Artificial Neural Network[J]. Journal of Remote Sensing, 2002, (1): 40-44. DOI： 10.11834/jrs.20020108.

摘要

介绍了一种基于人工神经网络的海中一类水域叶绿素反演方法。人工神经网络是 3层的反向传输神经网络。其结构是输入层有 4个节点

它们分别对应 4个波段 412nm

44 3nm

490nm

5 10nm的遥感反射比与 5 5 5nm波段遥感反射比的比值

隐含层有 5个节点

输出层一个节点对应于叶绿素浓度。该神经网络的训练和试验样本集来自SeaBAM的数据集。数据集中的 919个站位的 70 % （6 44个）用于训练

30 % （2 75个）用于测试。结果表明

该方法的精度优于被广泛采用的三次经验方法。

Abstract

A method to retrieve the oceanic chlorophyll concentrations in Casse I water based on Aritificial Neural Network （ANN） is presented. The ANN used in this paper is a three

ayer feed forward back

ropagation network which has 4 neurons in the input layer （corresponding to the ratios of the remote sensing reflectances at 4 wavelengths: 412 nm

443 nm

490 nm

510 nm to the remote sensing reflectance at 555 nm）

5 neurons in the hidden layer and one neuron for the output layer （Corresponding to chlorophyll concentration）. The training data set and testing data set of the ANN come from SeaBAM data base. 70% of the 919 stations in SeaBAM was used for training data set

the other 30% used for testing data set. At the end

the retrieved results from ANN and from the empirical algorithms were compared. The results show that the accuracy of the ANN is better than the cubic empirical algorithms used widely.