WANGs6019501图像处理作业.docx
《WANGs6019501图像处理作业.docx》由会员分享,可在线阅读,更多相关《WANGs6019501图像处理作业.docx(16页珍藏版)》请在冰点文库上搜索。
WANGs6019501图像处理作业
Question1:
Selecttheimagesyouwanttousefromthedataset.PerformafuzzyclassificationonselectedLANDSAT8OLIimagesusingR.Storescreenshotsofyourclassifications(waterclassonly)inanMSWorddocument.Specifyandmotivateyourchoicesoftheimagestoinclude,ofthenumberofclassesandoftheparameterm.Rememberthattheaimoftheclassificationistoseechangesinthesurfaceareaofthelakethroughtime.
First,Ineedtoselecttheappropriateimagestodotheclassification.Theprincipleofmyselectionisthattryingtoselecttheimageswhichhaveaslesscloudinthelakeareaaspossibleandthewaterareashouldbeasclearaspossible.Andthedateofimagesshouldbedifferent.Finally,Iselectfiveimagestodothefollowingwork.Theyarelc81690602013198.img,lc81690602013182.img,lc81690602013168.img,lc81690602013150.imgandlc81690602013246.img,separately.
TheimagesIselected
ThenIneedtodotheclassificationoftheseimages,respectively.ThefirstthingIneedtodoisdeterminingtheparameterswhichneedfortheclassifications.
Ncl—thenumberofclasses.ForeachimagesItrydifferentNclwhichare3,4and5,respectively.AndthenIchoosethefinalNclastheonewhichmakethewaterareamostclearlyaftertheclassification.Finally,theNclis5forlc81690602013198.img,lc81690602013182.img,lc81690602013168.img,lc81690602013150.imgand4forlc81690602013246.img.
m—theparameterwhichrepresenttheleveloffuzziness.Therangeofparametermis[1,infinite).Thefuzzinesswillincreasedasmincrease.Ifmequalto1,thefuzzyclassificationwillturntoacrispclassification.Itryseveralvaluesofm,andknowingifthemsettingtotoolarge,theiterationwillnotworkwell,themeanofclasseswillhardtoseparate.Andtheconfidencevaluewillbelow.ThenIsearchforsomearticlesaboutit,itseemsthatmequalto2willbetheempiricalbestvalue.SofinallyIselectthemequalto2.
Aftertheiterationofclassification,Ineedtoidentifywhichclassiswater.Asknowingthereflectionspectrumofthewater,thewaterhasthelowestreflectanceinband4-6,SoIcanidentifythewaterclassbycode:
Andtheresultoffuzzyclassificationsshowsbelow,intheseimagesthepixelvaluesarethemembershipvaluesofwaterclass.
lc81690602013150.img
lc81690602013168.img
lc81690602013182.img
lc81690602013198.img
lc81690602013246.img
Question2:
Foreachofthemembershiptowaterimagesyouobtained,groupthepixelsbelongingtothelakeintoalakeobject.IncludescreenshotsofthelakeobjectsinyourWorddocument.Explainandmotivatehowyoumadetheobjects.
AsIneedtoseechangesinthesurfaceareaofthelakethroughtime,it’sbettertocreatethelakeobjectsfortheimages.Thenit’sconvenienttoseethechangesbyobjectsovertime.
Thefirstthingtogroupthelakeobjectissettingthethresholdvalue
thr—thevalueofthreshold.Itdeterminesthatweatherthepixelcouldbeapartofthelakeobjectornotbycomparingitsmembershipwiththreshold.Isawthehistogramofmembershipofwaterclassandtriedseveralthresholdvaluestoseetheresultsofgroupingobjects.FinallyIchoose0.8asmythresholdtogroupthelakeobjects.
Whenaninitialpixelisdefined,itwillgrowin8directions,anypixelwhichadjacenttoitwillbejudgedbythreshold.Ifthemembershipvaluelargerthanthreshold,thisadjacentpixelwillbeincludedinthelakeobjectandwillgrowin8directionsagaindependonitself.Afterthegrowth,Iwillobtainmanysegmentsoflakeobjects,andthebiggestsegmentisLakeNaivasha.
FrompreviousexerciseIknowthelargestsegmentisLakeNaivasha,andthesecondlargestsegmentisLakeOloiden.Inthisassignment,IonlyrecordtheareasofLakeNaivashatoseethechangesthroughtime.Theresultofgroupinglakeobjectshowsblow:
lc81690602013150
lc81690602013168
lc81690602013182
lc81690602013198
lc81690602013246
Question3:
Calculatethesurfaceareaofthelakeforeachdate.ExplainhowyoucalculatedthelakesurfaceareaandshowtheresultforallLANDSAT8OLIimagesinatableandagraph.
ThesurfaceareaofLakeNaivashaisthesumofthenumberofpixelswhichinthelargestsegmentofthelakeobjectmultipliedbythepixelsize.ThesurfaceareasarecalculatedbyR.Theresultsare:
Thelakesurfaceareaeachdata
Originalimages
SurfaceareaofLakeNaivasha(m2)
lc81690602013150
125640900
lc81690602013168
129148200
lc81690602013182
134766900
lc81690602013198
127805400
lc81690602013246
135562500
Question4:
Evaluateyourresult.Didthesurfaceareachangeovertime?
Whatcanyousayabouttheuncertaintiesofthesurfaceareaestimatesandtheuncertaintiesinthechangesinsurfacearea.
Theuncertaintiesarisebothinthesurfaceareaestimatesandthechangesinsurfacearea.
First,theuncertaintiescomefromtheprocessing.Theparameterssettingfortheclassificationandgroupingobjecthastheinfluenceofuncertainties.Differentnumberofclassesordifferentparameterm,willleadtodifferentmembershipofallclasses.Andalsodifferentthresholdvaluewillleadtodifferentresultsoflakeobjects.Thatallwillleadtouncertaintiesonsurfaceareaestimatesandchangesinsurfacearea.
Then,theuncertaintiescomefromtheenvironment.FromtheimageIseecloudsshadowandalgainthelakearea.ThatwillmakeproblemswhenIdoclassificationtoidentifywater.Andleadtolessaccuracyonclassificationandgroupingobjects.
Question5:
Ifyoucomparethisresultwiththeresultoftheexercisesonfuzzyclassificationandobjectmonitoring,whereyouperformedsimilarproceduresonaseriesofLandsatETM+images,whatcanyousayaboutthedifferences?
ComparingtheseresultwiththepreviousexercisewhichIhavedonewiththelandsatETM+images.AsonlytwoimagesinLandsatETM+weretakenin2013andthedateare20130530and20130615,covertthesedatetoJuliandayare150and166.
Sothecorrespondingimagestopreviousexercisearelc81690602013150andlc81690602013168.
Thefuzzyclassificationandobjectmonitoringoftheseimages:
lc81690602013150vsl20130530
lc81690602013168vsl20130615
ComparingwithLandsatETM+images
images
Area
lc81690602013150
125640900
l20130530
125692200
lc81690602013168
129148200
l20130615
129022200
Andifyouseeonthegraph,itwillmoreclearly:
FromthetableandthegraphIseetheresultsfromthesetwodatasetsarequitesimilar.Buttherestillhavesomesmalldifference.Thatmaybebecausethespectralresolutionofthesetwodatasetaredifferent,andalsootherreasons.Forexamplethetimeoftakingthesepicturesarenotactuallythesame,andtheparametersIsettingtodoclassificationandobjectmonitoringarenottotallythesame.
Question6:
IfyoucomparetheNDVIplotshowwiththelakeareaplot,canyouseearelationshipbetweenthem?
Whichrelationshipdoyousee?
Didyouexpecttoseearelationshipbetweentheseplots?
Why(not)?
HistogramofNDVIperimageperpolygon
PlotofmeanNDVIofdifferentdate
ThechangeofsurfaceareaofthelakeandthechangeofmeanNDVIisrelevant.
Fromtheabovelinecharts,IseethechangeofmeanNDVIvaluesisalmosttheinversechangeoflakesurfacearea.Thatmeanwhenthesurfaceareaoflakeisincreased,themeanNDVIvalueisdecreased.ItistheexpectedrelationshipthatIwanttosee.Becausewhenthelakesurfaceareaisincreased,theareaofwaterincreased,theareaofplantsdecreased,thatleadtothedecreasingofNDVI.