遗传算法优化的BP神经网络城市不透水层百分比估算Estimating urban impervious surface percentage with BP neural network based on genetic algorithm
骆成凤
摘要(Abstract)：
本研究利用基于遗传算法优化的BP神经网络算法估算城市不透水层百分比。首先,将像元中各端元组分与BP神经网络的节点相对应进行BP网络建模,遗传算子建模;其次,对样本进行网络训练,先通过GA算法得到全局近最优网络权重集,然后用梯度下降算法训练网络,直到找到能充分反映特征空间中的数据分布模式的局部最优网络权重集;然后,训练好的网络被应用于整个影像用来估算城市不透水层覆盖百分比。在此基础上,对北京市地表不透水层百分比进行估算。试验结果表明,本研究所用的方法能有效利用中高分辨率遥感影像数据估算城市不透水层百分比。
关键词(KeyWords)： 城市不透水层;百分比估算;BP算法;遗传算法
基金项目(Foundation): 对地观测技术国家测绘局重点实验室开放基金项目资助
作者(Author): 骆成凤
DOI: 10.16251/j.cnki.1009-2307.2011.01.005
参考文献(References)：
- [1]Brabec E,Schulte S,Richards PL.Impervious Surfacesand Water Quality:a Reviewof Current Literature and ItsImplications for Watershed Planning[J].Journal ofPlanning Literature,2002,16(4):499-514.
- [2]Arnold J C A,Gibbons C J.Impervious Surface Cover-age:the Emergence of a Key Urban Environmental Indi-cator[J].Journal of the American Planning Associa-tion,1996,62(2):243-258.
- [3]Lu D S,Weng Q H.Use of Impervious Surface in UrbanLand-Use Classification[J].Remote Sensing of Envi-ronment,2006,102:146-160.
- [4]Yang X.Estimating Landscape Imperviousness Indexfrom Satellite Imagery[J].Geoscience and RemoteSensing Letters,IEEE,2006,3(1):6-9.
- [5]Wu C S,Murray A T.Estimating Impervious SurfaceDistribution by Spectral Mixture Analysis[J].RemoteSensing of Environment,2003,84:493-505.
- [6]Sangbum L,Lathro PR G.Subpixel Analysis of LandsatETM/sup+/Using Self-organizing Ma P(SOM)Neu-ral Networks for Urban Land Cover Characterization[J].IEEE Transactions on Geoscience and Remote Sensing,2006,44(6):1 642-1 654.
- [7]Lu D S,Weng QH.Spectral Mixture Analysis of the Ur-ban Landscape in Indianapolis with Landsat ETM+Im-agery[J].Photogram metric Engineering&RemoteSensing,2004,70(9):1053-1062.
- [8]廖明生,江利明,林珲,杨立民.基于CART集成学习的城市不透水层百分比遥感估算[J].武汉大学学报.信息科学版,2007,32(12).
- [9]周纪,陈云浩,张锦水,李京.北京城市不透水层覆盖度遥感估算[J].国土资源遥,2007,73(3).
- [10]江利明,廖明生,林珲,等.利用雷达干涉数据进行城市不透水层百分比估算[J].遥感学报,2008,12(1).
- [11]骆成凤,刘正军,王长耀,等.基于遗传算法优化的BP神经网络遥感数据分类算法[J].农业工程学报,2006,22(12).
- [12]Paola J D,R A Schowengerdt.A review and analysis ofbackpropagation neural networks for classification of remote-sensed multi-spectral imagery[J].International Journal ofRemote Sensing,1995a,16(16):3033-3058.
- [13]Gyanesh Chander,Brian Markham.Revised Landsat-5TM Radiometric Calibration Procedures and Postcalibra-tion Dynamic Ranges[J].IEEE Transactions on Ge-osience and Remote Sensing,2003,41(11).
- [14]赵英时.遥感应用分析原理与方法[M].北京:科学出版社,2003.