基于BP神經网络的遥感水深反演20110219.doc

Remote sensing of Extracting Water Depth based on Back Propagation Model基金项目： 基金项目：重庆市科委科技项目（CSTC，2009CB2015）； 作者简介：周廷刚黄山（1971 - ），男，四川省南部县人，博士，教授，主要从事生态环境遥感及相关问题研究。自己修改完善，E-mail: HYPERLINK mailto:zhoutg@163.com zhoutg@163.com 通信作者：周廷刚，博士，教授，E-mail: HYPERLINK mailto:zhoutg@163.com zhoutg@163.com Shan Huang1, , Tingang Zhou1,2,3 1) School of Geographical Sciences, Southwest University, Chongqing 2) Key Laboratory of Eco-environments in Three Gorges Reservoir Region, Ministry of Education, Chongqing, 3）Engineering Technology Research Center of Geo-spatial Information in Chongqing, Chongqing, China, 401147 Abstract: This paper aims to do water depth retrieval research in Lin huai gang segement of Huaihe river in Anhui Province. First, geometric calibration, atmospheric correction and water depth information extraction should be done to the remote sensing image acquired from Landsat 7 ETM+. Second, a water depth retrieval model was construct based on BP Model by using the correlation between measured water depth and the reflectance of the combination of single brand and multi-brand in the preprocessed remote sensing image. The result indicates that the mean absolute error, HYPERLINK /dict_result.aspx?searchword=%e5%b9%b3%e5%9d%87%e7%9b%b8%e5%af%b9%e8%af%af%e5%b7%aetjType=sentencestyle=t=average+relative+error average relative error, HYPERLINK /dict_result.aspx?searchword=%e5%86%b3%e5%ae%9a%e7%b3%bb%e6%95%b0tjType=sentencestyle=t=determination+coefficient determination coefficient and HYPERLINK /dict_result.aspx?searchword=%e5%9d%87%e6%96%b9%e6%a0%b9tjType=sentencestyle=t=root+mean+square root mean s square of the retrieved water depth and measured water depth are respectively0.5682m,12.11%,0.8465,0.7167m. 基于BP神经网络的遥感水深反演 黄山1 周廷刚1,2,3 1) 西南大学地理科学学院，重庆，中国 2) 三峡库区生态环境教育部重点实验室，重庆，中国 3）．重庆市地理空间信息工程技术研究中心，重庆，中国，401147 摘 要 本文在Landsat7 ETM+遥感影像预处理的基础上,利用影像单波段与多波段组合的反射率和实测水深值之间相关性，建立了安徽省淮河临淮岗段BP神经网络水深反演模型，进行了反演。结果表明：该模型反演水深值与实测水深值的平均绝对误差为为0.5682米，平局相对误差12.11%，决定系数