表格識別預(yù)處理技術(shù)與表格字符提取算法的研究.pdf

1、中山大學(xué)碩士學(xué)位論文表格識別預(yù)處理技術(shù)與表格字符提取算法的研究專業(yè)：計(jì)算數(shù)學(xué)學(xué)位申請人：謝亮導(dǎo)師及職稱：關(guān)履秦教授李小福副教授答辯委員會委員(簽名)766345／i們～噍$紅、力～影射誕茲翻秒擴(kuò)、＼，職歹堆符——薹苧塑至一——ResearchOnPreprOCessingandCharacterExtraCtionOfF0rmD0cumentRecOgnjtionMajo’NameSuDervjsorC0mDutatjOnaIMathe

2、ma“csxle。LIan9ProfessOrGuanLutaiAssociateProfessOrLi，XiaOfuAbst怕ct：OCRhasreceivedconsiderabIeattentionjnrecentdecadesAndfo咖documentrecognitjonhasbeenanimpOnant丘eldofOCRresearchandapplicatjonThjsdissenationmainlyfocuseson

3、twoissuesoffo冊documentrecognjtjon：pre—pmcessingandcha陽cterextractjonBinarizationaPpearstobeaIlimponamstepinimagepTocessingandcharacteTreco甜lition。110overcomethedraw南acksofcuⅡentmethodsinthe1iterature—noisesensitive，10wsp

4、eed，andarousingbrokenstmkesetc，ahybridmethodbasedonnon1inearcontrastellllancingaIldhG0peratorisprcsentedtobinarizefomdocumentjmagesExperimentsshowthatourapproachhassi印ifjcalltadVantagesSkewsareinevitablyintroducedduringt

5、heoauIseofsc砌ingAnewapproachbasedonHou91】t『柚sfonn—Hou曲t塒1sfo珊wjthparamcterconstraints，jsimplemenledinthes∞peoffO徹docIlmenlimageSInco|1】binationwithafasta彤net硎nsfQ瑚，this印pToachappearslodramaticallyspeedupskewcOrfectionOne

6、ofourcriticaltaSksistodealwiththesituationinwhichhand、】l硝ittencharactersareove『lappingbordersAnovelapproachispresentedtosolVethisproblem：Fo冊cellsareaccuralelylocatednⅡ’ouglIaCrFprocess知dcharactersoveflappingboI＆Ts缸eperfe

7、ctlyeX虹act討byutmziIlganECCEAmethodDe_noisingandsmoothingarecaIricdoutasalaststageExperimentsproVetheef!fectivenessofourapproach。Finally’withlhehelpofthemethodsaIldalgoriIhmsp『0posedjnthisdissertatjon，andwjIhtheex口loitati

8、onofoIlrlatcstreSearchinareaslikecharactersegmentation，fcatufcextracfjonanddassification，anautomaIictraIlscrjptreco印itionsystemisdeveloped，withareco鰓itionrateof85％K鼬，1II『Ords：oCR，f01皿documentIecogn“ion，biⅡaIization，skewc