资源描述:
《基于mean+shift的高分辨率遥感图像分割分析》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、浙江工业大学硕士学位论文RESEARCHONHIGHRESOLUTIONREMOTESENSINGIMAGESEGMENTATIONBASEDONMEANSHIFTABSTRACTTherapiddevelopmentofremotesensingtechnologyledtotheenhancementofremotelysenseddataacquisitiontechnology,allowinghigheraccuracyandlarger-scaleremotelysenseddatatobecomethemainobjectinthe
2、fieldofremotesensing.Therefore,themoretargetedtechniqueneedstobeprovidedbecausethetraditionalremotelysensedimagetechniquesarenotapplytothecurrenthighresolutionandlarge-scaleimage.Imagesegmentationtechnologyisakeystepintheprocessofremotesensingimageprocessingandprovidesconveni
3、encefortheoperationofimageclassification,featureidentificationandsoon,sotheaccuracyoftheresultisespeciallyimportantonthefinalimagequality.Inthispaper,thestudyonhighresolutionremotesensingimage,whichisbasedonMeanShiftalgorithmcombinedwiththefeaturesofremotesensedimage,hasbeenp
4、rocessing,andthemainworkandachievementsareasfollowed:1.IntroducethemainprincipleofMeanShiftalgorithm,includingtheselectionforcriticalpartanditsfeatures.Andintroducethebasicprocessofimagesegmentation,whichelaboratestheprocessandeffectofkeyprocessonMeanShift.2.Theoptimizedimpro
5、vementhasbeencarriedoutonsegmentationefficiencyandsegmentationaccuracyofhighresolutionremotesensingimagesegmentationbasedonMeanShiftalgorithm.Atfirst,anoptimizedalgorithmforparallelsegmentationoflarge-scaleimageonMeanShifthasbeenproposedtodealwithsomeproblemsoftraditionalsegm
6、entationtechnique,suchasthelowefficiencyandlowmemory,orevenimpossibletoseparate.Atthesametime,itcaneliminatethe“blockline”existedinblockparallelprocesstosomeextent,ii万方数据浙江工业大学硕士学位论文whichimprovesthesegmentationaccuracy.AndtheresultsiscomparedwitheCognitionsoftwaretoprovetheef
7、fectivenessofthisalgorithm.Secondly,makethestudyfordistancemetricinvolvedinmergeprocessofremotesensingimagesegmentation,whichreplacethetraditionEuclideandistancewiththesimilaritymetricinlinewiththefeatureofremotelysenseddata.Intheapplicationofhighresolutionremotesensingimages
8、egmentationonMeanShift,therelevantexperimentdatahasprovedthereplaced