基于上下文的機(jī)場目標(biāo)識(shí)別方法.pdf

1、華中科技大學(xué)碩士學(xué)位論文基于上下文的機(jī)場目標(biāo)識(shí)別方法姓名：劉蒽蒽申請學(xué)位級別：碩士專業(yè)：模式識(shí)別與智能系統(tǒng)指導(dǎo)教師：桑農(nóng)2003.5.9華中科技大學(xué)碩士學(xué)位論文●■●●_■■■●____■●■■●■■__■■■■●_IllI_●■●■■■■■__■■■__■■■■■__●■■■___■■■___■■__|_■■___■●■_■●■■■■●■■■■■●■■__●■●一AbstractTheAutomaticTargetRecognitio

2、n(ATg)ofremotesensingimagesisoneofthemostimportantresearchprojectsintheATRdomainsWiththedevelopmentofChinesespaceandremotesensing，it’SnecessarytorecognizetheinterestedtargetsfromimagesautomaticallyTheobjectsonthegroundar

3、einspecificenvironmentAsatypicalmulticomponentclassofobjects，airportsandtheircomponentscontainmuchcontextualinformationTherefore，anapproachbasedoncontextualinformationisproposedtecognizeairportsinthisthesisTheapproachwit

4、hcontextwillenhancethevalidityofrecognitionFirstlythisthesisanalyzesthestructureoftheairportsandgetssomefeaturesoftheseobjectsintheremotesensingimages，whichwillbethebasisofrecognitionThesefeaturesinclude：parallelrelation

5、shipbetweenrunways；perpendicularrelationshipbetweenexittaxiwaysandrunways；theamountsofexittaxiways；andthegrayvaluerelationshipbetweenthebackgroundandthetargetssuchasrunways，taxiways，apronsandexittaxiwaysSecondly’theairpo

6、rtsandtheircomponentshaveobviousorientationInthisthesis，theoriginalimageisdecomposedintomultiplespatialfrequencyandorientationselectivechannelsbyGaborfiltersEachchannelproducesonefeatureimageThirdlyRadontransformisadopte

7、dtoextractlinefeaturesfromeachfeatureimageThecandidaterunwaysareverifiedfortheirparallelism，andtheobviousfalserunwaysareeliminatedThen，thepotentialtargetareaisdelimitedThecandidateexittaxiwaysareextractedfromthisareaandv

8、erifiedbytheir掣ayvalueThuswewillgetcreditablerunwaysandexittaxiwaysFinallyaccordingtothecontextualinformationofairports，weuseaseriesofcriteriatodeterminetheairportintheimageTheairportsrecognitionapproachwithcontextisshow