图像五值化和基于三值图像车牌识别

《图像五值化和基于三值图像车牌识别》由会员上传分享，免费在线阅读，更多相关内容在工程资料-天天文库。

1、JournalofSoftware软件学报图像五值化与基于三值图像的车牌识别张忠义1+1(海口愚佬会教育科技有限公司,海南海口570203)Image5-Value-SegmentationandLicensePlateRecognitionbasedon3-valueimage*ZHANGZhong-Yi1+1(HaikouYulaohuiEducationTechnologyCo.,Ltd,Haikou570203,China)+Correspondingauthor:Phn:+86-898-36393679,Fax:+86-898-36393679,E-ma

2、il:yulaohui@qq.comReceived;AcceptedAbstract:Image5-value-segmentationisaimagesimplificationmethodusedtomapRGBspaceintoYBspace,bytheformulationY=(R+G)/2andthenconvertcolorinYBspacetofivevalues:black、blue、gray、yellowandwhite.ForChineselicenseplaterecognition,fivevaluescanfurtherbecompres

3、sedtothreevaluesbycombiningblackandblue,whiteandyellowtogetherseparately.Themethodproposedhastwoapplicationscenarios:characterrecognitionbasedon3-valueimageandlicenseplatelocationbyinversingsegmentationinformationbasedon3-valueimage.Inthefirstscenario,Thereservedgrayin3-valueimagecanbe

4、classifiedintoeithertheplatebackgroundcolorortheforegroundcharactercolorforthecharacterrecognitionaccordingtothelicenseplatebackgroundcolorrecognized.Itcanmakeupfortheinformationlossintheprocessofimagebinarizationbydeterminingthreshold,whichisalwaysinfluencedbytheunevenilluminationordi

5、rtyspots.Doingthishelpstoimprovetheaccuracyofcharacterrecognition.Inthesecondscenario,candidatelocationoflicenseplatecanbefoundinthepremiseoftraversingtheimageefficientlyifcharactersegmentationcanbeperformedbasedonprojectionsofthe3-valueimage.Thismethodcandetectcharacterspacingfirstlyt

6、omeettheneedsofusers’andeliminatefalselicenseplate，makesuretheaccuracyoflicenseplatelocationwhichdoesn’tdependonlicenseplatecharacter;itcanalsoreducethetruenegativebyignoringtheinferencebroughtbyplatefeatures.Keywords:5-valuesegmentation;licenseplaterecognition摘要:图像五值化是通过公式Y＝（R＋G）／2，将R

7、GB空间映射到YB空间，再将YB空间简化成黑、蓝、灰、黄、白五种值的图像简化方法。针对中国车牌，通过黑、蓝合并，白、黄合并，图像可进一步简化成三值。基于三值图像的车牌识别包括基于三值图像的字符识别和基于三值图像按分割反推的车牌定位。三值图像保留的中间灰色可根据识别的车牌底色视同字符识别所需的车牌底色或字符色，从而能弥补受光照不均匀或车牌“脏”影响按阈值选取进行图像二值化可能导致的图像信息缺失，有利于提高字符识别的正确率。基于三值图像按分割反推的车牌定位是指在确保图像遍历高效的前提下，通过基于三值图像的投影如果可以进行字符分割则得到一个候选车牌定位。其好处是可以先