第32卷第3期测绘工程Vol.32No.32023年5月EngineeringofSurveyingandMappingMay2023DOI:10.19349/j.cnki.issn1006-7949.2023.03.010多源遥感数据支持下的城市不透水面提取———以天津市为例张泉(天津市测绘院有限公司,天津300381)摘要:近年城市人工地表的快速扩张引发了一系列生态问题,准确高效地获取城市不透水面信息及空间分布将具有重大的意义。以天津市为实验区,融合哨兵一号、哨兵二号及高程数据,构建多特征组合的随机森林模型,实现基于高空间分辨率影像的不透水层信息识别,并与其他方法进行比较。结果表明,利用本方法提取天津地区不透水面信息的总体分类精度达90.93%,Kappa系数达0.877,与其他方法的对比实验也进一步证明本方法在抑制错分误差和漏分误差方面达到了较好的效果,为未来实现大范围高精度不透水面制图奠定基础,为城市发展和生态保护提供有力依据。关键词:城市不透水面提取;随机森林;多源遥感数据;特征提取中图分类号:TP79文献标识码:A文章编号:1006-7949(2023)03-0069-07Extractionofurbanimpervioussurfacesupportedbymulti-sourceremotesensingdata:acasestudyofTianjinZHANGQuan(TianjinInstituteofSurveyingandMappingCo.,Ltd.,Tianjin300381,China)Abstract:Therapidexpansionofurbanartificialsurfacesinrecentyearshascausedaseriesofecologicalproblems.Itwillbeofgreatsignificancetoaccuratelyandefficientlyobtainurbanimpervioussurfaceinformationandspatialdistribution.ThispapertakesTianjinastheexperimentalarea,Sentinel-1,Sentinel-2andelevationdatawereintegratedtoconstructamulti-parameterrandomforestmodel,whichrealizestheextractionofurbanimpervioussurfacebasedonhighspatialresolutionimages,andcompareswithothermethods.TheresultsshowthattheoverallclassificationaccuracyoftheimpervioussurfaceextractedbythismethodinTianjinisupto90.93%,andtheKappacoefficientisupto0.877.Thecomparativeexperimentwithothermethodsalsofurtherprovesthatthemethodinthispaperhasachievedbetterperformanceinsuppressingmisclassificationandomissionerrors,whichlaysthefoundationfortherealizationoflarge-scaleandhigh-precisionimpervioussurfacemappinginthefuture,andprovidesastrongbasisforurbandevelopmentandecologicalprotection.Keywords:urbanimpervioussurfaceextraction;randomforest;mu...
