摄像机运动条件下动目标检测方法分析-analysis of moving target detection method under camera motion.docxVIP

AbstractAsoneoftheimportantcontentsofcomputervision,movingobjectdetectionapproachisthebasisofobjectdetectionandtracking,involvingalotofknowledge,suchasimageprocessing,patternrecognitionandartificialintelligence,andautomaticcontrol.Ithaswideapplicationprospectinsecurity-monitoring,intelligenttraffic,nationaldefenseandmilitary,medicalandsoon.Withthefurtherdevelopmentandwidelyusedofmovingobjectdetectiontechnique,andpeopleputforwardmorerequirementstowardsit,thetendencyofmovingobjectdetectionapproachfromthefixedcameraconditiontothecameramotionconditionisinevitable,intheresearchofcomputervision.Thispaperintroducesakindofmovingobjectdetectionapproachintheconditionofcameramotion.Mainresearchcontents,results,andinnovationsareasfollows:Motionestimationalgorithmintheconditionofcameramotionisstudiedindetail.Forvideomotion,thebackgroundimageisconstantlychanged,whichisseriouslyaffectedtheaccuracyofmovingobjectdetection.Soastosolvethisproblem,combinedwithmotionestimationandmotioncompensationtechnique,motionestimationalgorithmbasedonnon-motionregionpartitionsisproposedinthispaper.CornersareextractedbyHarrisinthereferenceimage,andallinmotionregionpartitionareremoved.Then,pixelpointsarematchedwithcornersinthenon-motionregionpartitionofthereferenceimage,andmotionvectoriscomputedtocompensatethecurrentimage.Manygroupsofexperimentalresultsshowthatthealgorithmcanhavehighlevelofprecisionandexecutiveefficiency.Motionobjectdetectionapproachintheconditionofcameramotionisstudiedindetail.Onbasisofmotionestimation,andcombinedwithconnectedregiontechnology,thispaperproposesakindofmotionobjectdetectionapproachbasedonedgeinformationandconnectedregion.ThismethodusesOtsu’smethodtogetthethresholdandconvertsthedifferenceimagewithbinaryprocessing.Then,K-nearestneighbormedianfiltermethodandmorphologicalfiltermethodareusedtofilterinterferencenoisesintheimage.Last,Movingobjectisdetectedbyedgeinformationandconnectedregionintheimage.Manygroupsofexperimentalresultsshowthattheapproachcanaccuratelydetectmovingob