運動分割的改進算法及其在智能交通系統中的應用.pdf

1、華南理工大學碩士學位論文運動分割的改進算法及其在智能交通系統中的應用姓名：李穎雄申請學位級別：碩士專業(yè)：信號與信息處理指導教師：孫季豐20040601AbstractMPEG。。4andMPEG。。7appearsasthesecondgenerationofvideocompressionstandardSegmentation—basedcoding，objectbasedcodingandmodel—basedcodingaret

2、hreeofmaintechniquesofMPEG4ThisthesisdiscusesmotionsegmentationanditsapplicationinintelligenttransportationsystemAndsimulationexperimentisperformedonthecomputerwhichrealizethecompleteprocessofmotionsegmentation，including

3、attainmentofvehicleimagesequence，globalmotioncompensation，opticalflowfieldestimation，movingobjectsegmentationandtrackingWhenwedotheglobalmotionestimationofimagesequenceafastestimationalgorithmbasedon4parametermodelisprop

4、osedWefirstlyuseanapproximateformulatocalculatethedifferentialcoefficient，theestimatethe4parameterthroughstatisticaltheoryandthresholdfilterThenwegeneratethecompensateimagebybilinearinterpolation。Alayeredandmulti—thresho

5、ldBMA(Block—MatchingAlgorithm)isusedtodomotionestimationThealgorithmisdividedintolayersthroughdifferentsearchingsteplengthIneachlayerweusedifferentsearchtacticsandmatchcriterionAndathresholdissettobreakdownthesearchingSO

6、astodecreasethecomputationHowtodividethelayersandsetthethresholdisdecidedbythemovingobjectsandmotionExperimentalresultshowstheproposedalgorithmisbetterthanthetraditionalBMAobviouslyatoperationspeedandeffectThemotionsegme

7、ntationalgorithmWeusediscomposedoftwoparts，thatisBMAmotionestimationandthechangedetectionFirstlyweusea毋obalthresholdandmathematicalmorphologyfiltertodeterminethechangingregionThenaccordingtOtheresultofmotionestimation，we

8、judgethemovingobjectregionincludingpixelswhosemotionjumping—offpointandend—pointbothbelongtothechangingregionExperimentresultprovesthatthealgorithmcansegmentthemovingvehiclefromtheimagesequencesuccessfullyFinallyasimplem

9、otionobjecttrackingalgorithmisproposedWeusethechangedetectionresultoftheimagesequencetorecoverthebackgroundimageThenthesegmentedmovingobjectiscasttothebackgrounddependsonthevelocitywhichcanbecalculatedbymotionestimationW