基于改进的lm-bp神经网络的车牌识别算法研究与应用

《基于改进的lm-bp神经网络的车牌识别算法研究与应用》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、广东工业大学硕士学位论文ABSTRACT·Licenseplaterecognitionsystemisanimportantpartofintelligenttransportationsystem，anditisalsoahotspotofdomesticandforeignscholars’research．Bypreprocessingoflicenseplatelocationandextraction，thelicenseplateimageprocessing，licenseplatefeatureextraction，licenseplatecharacte

2、rrecognitiontechnologytoidentifyvehiclesgrades，colorandinformation．Italsocombinesmanytechniquessuchasimageprocessingandpatternrecognition，whichCanbewidelyusedinthefieldofhighwayautomatictollcollection，roadmonitoring，communityandvehiclemanagementsystem．Inthepracticalapplication，licenseplate

3、recognitionstillexistmanybottlenecks，suchasrainyday,fogandothercomplexenvironmentunderlowrecognitionrateandlowspeed；eventhelicenseplaterecognitionisnotidealinsituationofexposuredegreeistoolowortoohigh，etc．Inordertosolvetheseproblems，thispapermakesadeepstudyoflicenseplatecharactersegmentati

4、onandcharacterrecognition，putsforwardsomeimprovedalgorithm，andUSeStheVC++6．0platformtodevelopthelicenseplaterecognitionsysteminthisthesisandtherelatedexperimentsarecarriedout．．Inthispaper,avarietyoflicenseplaterecognitionalgorithmbasedontheanalysis，andbasedonthisproposedthecorrespondingimp

5、rovedalgorithm，whilethelicenseplateimagepreprocessing，flitimagecorrectionandothercontent．Inordertoeliminatethepollution，improvethebinarizationaccuracy,thispaperpresentsbinarizationalgorithmbasedonthevirusevolutionarygenetic，whichisakindofimprovedgeneticalgorithm，inordertomaintainthediversi

6、tyofpopulationandthevirusmechanismcombinedwiththegeneticalgorithm．TheadaptivethresholdCannyedgedetectionalgorithmforedgedetection,andmatchtheRadontransformmethod，inthedetectiontotheedgeoftheinformationtoaccuratelydrawthelicenseplatetiltangle，thecorrectioninordergetthecorrectimage．Onthisbas

7、is，inordertoreducetheblurringofthelicenseplateimagesegmentationerrorrate，thisPaperimprovetheverticalprojectionofthesegmentationmethod．Inordertoimprovethecomplexenvironment(rainy,foggyanddim，etc．)ofthecharacterrecognitionrate，thispaperpresentsakindofimprovedLM—