車牌字符分割方法研究與實(shí)現(xiàn).pdf

1、西南交通大學(xué)碩士學(xué)位論文車牌字符分割方法研究與實(shí)現(xiàn)姓名：張學(xué)海申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：計(jì)算機(jī)應(yīng)用技術(shù)指導(dǎo)教師：彭強(qiáng)20100501AbstractLicensePlaterecognitionsystemisoneimportanttopicoftheIntelligenceTra銜cManagementLicenseplaterecognition(LPR)hasgreatapplicationprospectiveintolling

2、，monitoringtrafficflowinvestigatingon—the—spottrafficaccident，handlingbreaksofthetrafficregulationsandkeepingparkingsecurePeoplepayincreasingattentiontoLPRbasedondigitalimageprocessingandpatternrecognitionLicensePlaterec

3、ognitionsystemisanimportantsubjecttobestudiedintheintelligenttrafficItismainlycomposedofthreekeyparts：licenseplatelocationlicenseplatecharactersegmentationand1icenseplatecharacterrecognitionLicenseplatecharactersegmentat

4、ion，asakeypartofthelicenseplaterecognitionsystem，ismainlystudiedinthisPaperTherearethreekeypointsinthelicenseplatesegmentationsystemTheyarebinarization，tiltcorrectionandcharactersegmentationInthebinaryalgorithmthethesish

5、ascombinedwithadvantagesofthetwoglobalbinarizationOtsualgorithmandpartialbinarizationBgrnseninordertoadoptanewmethodofthresholdsegmentationwithlicenseplatefeatures—FourSectionBlockBinaryAlgorithmInthetiltcorrectionalgori

6、thm，thethesishasadoptedthetiltcorrectionalgorithm，combiningwithlinearfittingandRadonChangeforhorizontaltiltcorrectionHoweverforverticaltiltcorrectionthethesishasappliedcharacterprojectiontogetverticaltiltangleandcalibrat

7、edthetitlebythemethodofcoordinatetransformationItCanavoidlosingmuchcharacterdetails，behelpfultoconsequencesegmentationandCanimproveidentificationrateInthecharactersegmentationalgorithm，thethesisadoptedamethodof1icensepla

8、techaractersegmentationwithconnectivityverticalprojectionandtemplatematchingFirstlythecharacter’Swidthhasbeenfoundby也econnectivity；Secondlyseparatorlocationhasbeenfound，accordingtotheprojection；Atlast，thesegmentationhasb

9、eencarriedbythetemplatewhichhasbeensetupbyconnectivitycharacter’SwidthThen，singlecharacterhasbeenadjustedmicroscopicallybyprojectionTheauthorproposesthecorrespondingsegmentationmethodforeachtypeoflicenseplate，afterconsid

10、eringfullyexistinglicenseplatetypesespeciallyfordoublelicenseplate，policeandconsularlicenseplatesetcLiceriseplatesegmentationsystemiSselfadaptiveItcouldprocesstheimagesthatsampledfromdifferentenvironmentandgetawonderfulr