摘要
基于径向基函数(RBF)神经网络技术设计了一种新的超光谱热红外大气辐射传输模型快速算法,通过模拟实验,确定神经网络样本和输入输出数据,并对典型波段训练出相应的多层神经网络,用于快速计算超光谱热红外大气顶部辐射传输亮度。实验中分别训练了9μm、10μm和12μm波长处相应的神经网络,实验结果表明所建算法不仅具有较高的计算精度,而且每个波长对应亮温的计算速度比利用4A模型的计算速度最多可提高100倍以上,同时,在波段的选择上也具有更高的灵活性。
Surface temperature and surface emmisivity are two important parameters for earth environmental researches.Theoretically they can be retrieved with radiance data received from satellite,but in practice,more constraint conditions are needed to solve this problem.Taking advantage of the large amount of channels,hyperspectral remote sensing data provides a new way for this problem.In order to develop a new model to separate surface temperature and surface emmisivity fast and accurately with the help of hyperspectral remote sensing data,a fast method to assimilate the radiance data received by hyperspectral sensors such as Infrared Atmospheric Sounding Interferometer（IASI） must be developed at first.Currently,some kind of hyperspectral infrared atmospheric radiative transfer model（RTM） have been applied in numerical weather prediction（NWP） systems.Though these models have highly accuracy,they still can not meet the requirement for calculation speed.This paper developed a fast calculation approach for the hyperspectral infrared radiative transfer model based on artificial neural network,which would significantly improve the calculation speed of RTM and at the same time be relatively accurate compared with other hyperspectral infrared atmospheric RTMs.In recent years,Artificial Neural Network（ANN） has been combined into some atmospheric RTMs,such as NeuroFlux in ECMWF＇s atmospheric model and NN emulation in NCAR CAM longwave atmospheric radiation parameterization.With the help of ANN technique,these models could enjoy a highly calculation speed for radiance and other related parameters.However,existing methods could not be suitable for the fast calculation of hyperspectral models.Automatized Atmospheric Absorption Atlas（4A） is an accurate hyperspectral infrared radiative transfer model which is suitable for the simulation of hyperspectral infrared thermal sensors such as IASI.But it sill takes too much time for model calculation.In the paper,radiance and atmospheric parameters calculated with 4A model were used as true value to judge the accuracy of fast calculation approach,and the calculation speed of 4A would be compared with fast approach too.In this paper,RBF（Radial basis Function） neural network technique is introduced to design a fast calculation approach which is used to accelerate the calculation speed of hyperspectral infrared thermal radiative transfer model.The proper inputs and outputs for our proposed neural network have been found through a lot of numerical simulations and calculations.In addition,a type of multi-layer neural network structure has also been developed to fast calculate the top of atmosphere radiance for typical wavelengths in hyperspectral thermal infrared spectrum.In this paper,three sets of neural networks were trained respectively for three wavelengths：9,10 and 12 μm.Results of the experiment show that this fast calculation approach could calculate the top of atmosphere radiance with the error less than 0.1 K compared with 4 A,and its running speed is 100 times faster than that of 4 A for a single wavelength.This approach also enjoys more flexibility for the choosing of spectral channels.
出处
《干旱区地理》
CSCD
北大核心
2010年第1期99-105,共7页
Arid Land Geography
基金
国家863计划项目(2006AA12Z121
2006AA12Z148)
中国科学院知识创新工程重要方向项目(KZCX2-YW-Q10-2)资助
关键词
大气辐射传输
径向基函数神经网络
超光谱热红外遥感
atmospheric radiative transfer RBF neural network hyperspectral infrared thermal remote sensing
