Geometric Correction for Images of Intersection Taken by Hovering UAV

Low altitudeunmannedaerialvehicle(UAV)imagesshaveincreasinglydemonstrateduniquead vantagesinthefieldsoftransportplanning,urbanmanagement,emergencymonitoring.TounderstandtheadvantagesofUAVimages,rapidandaccuratecorrectionandregistrationoftheimageareanecessaryprem ise.ThispapertakesaimageofintersectiontakenbyhoveringUAVasitsresearchtarget,anddiscussesthemethodofgeometriccorrectionandregistrationforsuchimages.Theimagecontainsalargenumberofsimilarfeaturesandmovingobjects,whichmakesitdifficulttobecorrectedintraditionalway,basedonspatialdo mainorfrequencydomain.Thus,thispaperproposesanimageregisteringmethodbasedonsemanticinforma tionextractedfromzebralines.Torealizeimagesregistrationbetweenarbitraryframes,thefollowingstepsarenecessary:patternsextraction,zebralinepatternspreparationandclassification,axisfittingandimageregistration.Finally,atestisperformedwithavideotakenbyUAV.Theresultsdemonstratethatthemeth odisrobustandeffective.