人工神经网络算法在红外高光谱资料反演大气温度廓线中的应用.pdfVIP

33 3 V o.l 33 N o. 3 2010 6 T ransactions o f A tm ospheric Sciences Jun. 2010 , , . [ J]. , 20 10, 33 3) : 34 1346. Guan Li, L iu Y ang, Zhang Xu ehu i. A ppl icat ion of art if icial n eural network algorithm in retrieving atmosph eric temperature prof iles from hyperspectral in frared data[ J]. T ran s A tm os Sci, 20 10, 33 3 ): 34 1346. 官莉, 刘旸, 张雪慧 ( , 2 10044 ) : 基于红外高光谱大气探测器A IR S 实况观测资料, 尝试用人工神经网络算法来实现晴 时大 气的温度垂直廓线反演, 重点将该算法与目前已经发展比较成熟的特征向量统计反演算法进行比 较结果表明, 两种算法在计算时间效率和反演精度上相当鉴于人工神经网络算法的统计物理 本质, 误差反向传播 BP 神经网络可以求解非线性问题, 在优化训练样本和继续调试网络种类和网 络参数的基础上, 有望能进 一步提高反演精度 : 红外; 高光谱; 人工神经网络; 反演 : P407. 6; TP79 : A : 2010) Application ofArtificialNeuralNetwork Algorithm in Retrieving AtmosphericTemperatureProfiles from Hyperspectral Infrared ata GUAN L ,i L IU Y ang, ZHANG Xu ehu i K ey Laboratory of M eteorological D isaster of M inistry of E du cation, NU IST, N anjing 2 10044, Ch ina) Abstract: T he art ific ial neural netw ork algo rithm is presented in th is paper to retrieve the tem perature pro files under clear sk ies by using A IR S A tm ospheric InfraRed Sounder) actual observat ions. T he study is focused on the com parison of art ific ia l neural newt ork retrieval algo rithm w ith eigenvector regression al gorithm wh ich has already been w e ll deve loped. T he results show that these tw o a lgorithm s cost nearly the sam e com put ing tmi e w ith com para tive prec ision in the real A IR S data retr ieval process. Due to the statis t icphysical nature o f the art ific ial neural n