引用格式：林娜,何静,王斌,等.结合植被光谱特征与Sep-UNet的城市植被信息智能提取方法[J].地球信息科学学报,2023,25(8):1717-1729.[LinN,HeJ,WangB,etal.IntelligentextractionofurbanvegetationinformationbasedonvegetationspectralsignatureandSep-UNet[J].JournalofGeo-informationScience,2023,25(8).1717-1729.]DOI:10.12082/dqxxkx.2023.220866结合植被光谱特征与Sep-UNet的城市植被信息智能提取方法林娜1，何静1*，王斌2，唐菲菲1，周俊宇1，郭江11.重庆交通大学智慧城市学院，重庆400074；2.重庆市地理信息与遥感应用中心，重庆401147IntelligentExtractionofUrbanVegetationInformationbasedonVegetationSpectralSignatureandSep-UNetLINNa1,HEJing1*,WANGBin2,TANGFeifei1,ZHOUJunyu1,GUOJiang11.SchoolofSmartCity,ChongqingJiaotongUniversity,Chongqing400074,China;2.ChongqingGeomaticsandRemoteSensingCenter,Chongqing401147,ChinaAbstract:Urbanvegetationisanimportantcomponentofurbanecosystemsandplaysavitalroleinhumansettlements,urbanecology,urbanplanning,andsustainabledevelopment.Itisurgenttodevelopanefficientandaccuratemethodtoachievetheintelligentextractionofurbanvegetation.Inviewoftheproblemsoflowefficiencyandstronghumaninterventionintheextractionofurbanvegetationbytraditionalmethods,andinsufficientutilizationofspectralinformationindeeplearningmethods,thisstudyfocusedonintelligentextractionofurbanvegetationfromGF-1DhighresolutionremotesensingimagesbycombiningthemostimportantspectralreflectioncharacteristicsofvegetationintheNearInfrared(NIR)bandandtheSepU-Net,anoptimizationofU-Net.Themainworkofthisresearchincludes:(1)wecreatedthreesamplesetsconsideringthehighreflectanceofvegetationintheNIRband:thetrue-colorgreenspacesampleset(truesampleset),whichservedasthecontrolgroup,thestandardfalse-colorgreenspacesampleset(thefakesampleset),andthefalse-colorgreenspacesamplesetsynthesizedbyNDVI(NDVIsampleset);(2)theSep-UNetwasoptimizedbasedonU-Net.OnthebasisofU-Net,Sep-UNetexpandedthenetworkreceptivefieldandincreasedthenetworkdepthbyincreasingthenumberofconcatenatedconvolutionstoachievethepurposeofenhancingthenetwork'sinforma...
