定场景内随机运动车辆检测分类算法研究

《定场景内随机运动车辆检测分类算法研究》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、中国科学技术大学硕士学位论文定场景内随机运动车辆检测分类算法研究姓名：郭利生申请学位级别：硕士专业：电路与系统指导教师：郭立20060501摘要宽、高、紧致性等)和二维不变矩特征，它可以作为定视角下图像采集器和运动车辆空间关系发生变化时的车辆分类特征。4．提出一种基于定视角下三维不变量特征的随机运动车辆分类算法当前基于图像的车辆分类算法只能区分车辆和行人以及大型车和小型车，特别是当图像采集器与运动车辆的空间关系发生变化时，不能对车辆进一步细分，本文提出了一种适用于固定视角下的基于Radon不变矩特征曲线的分类

2、算法。实验结果表明，对随机运动中的八类车辆的平均分类准确率为96％。关键字：运动检测阴影检测RTS不变量车辆分类ABSTRACTComputerVisionSurveillanceispopularnow．WiththedevelopmentofInfclnnationTechnologyandthetheoryofcomputervisionsurveillance．DigitalVisionSurveillancesystemreplacetraditionalsimulatesurveillancesy

3、stemstepbystep．Now．AutomaticVehicleClassificationsystemsarewidelyusedinnationaldefenseandciviliansystemBut．theexitingalgorithmsforAVCCarlonlysuitforself-definitioncondition，thedependenceonself-definitionconditionmadethiskindalgoritbanscan’tbewidelyused，andi

4、ncapableofClassificationofrandommovingvehicles．Butthisabilityisthemostimportantoneinbattlefieldsurveillanceand10cationsurveillance．RandomMovingVehiclesClassificationsystemincludesseveralstages：motiondetection．movingshadoweliminationandtargetrecognition．Toac

5、hievealealtimeandhighlyadaptabilitysystemthatdidn’tdependonthecharacterofimageandcamerapositionItmustbeensuredtllatallstagesofsystemarecapableofit．Mostofexitingalgorithmsimplythatitisfixedtoself-definitioncondition．Itisnecessarytodevelopthealgorithmswithrea

6、ltimeandhi【ghlyadaptabilityt11atdidn’tdependonthecharacterofimageandcamerapositionforAVC．The血esisfocusesonRandomMovingVehiclesClassi矗cationalgorithms．Themaininnovationsarelistedasfollows．1．AimprovedmovingobjectextractionalgorithmbasedDesaSMmixedimagediffere

7、nceisproposed．In2004，DesaS．Mproposedamixedimagedifferenceforextractingmovingobjectfromstaticirrelevantbackground．Theyhaveobtainedmotionmaskbyapplyingthebackgroundsubtractionandconsecutiveframedifferencing．Forallcomputebasedfullimage．thereisinevitablyalotofr

8、ed．dance．Region—basedtechniquewasemployedinproposedalgorithmtocutdowntheredundance．adynamicfactortoeliminatemovingobjectspeedeffectonconsecutiveframedifferencingTheexperimentalresultsdemonstratethatthe