資源描述:
《數(shù)字字符檢測(cè)與識(shí)別方法研究》由會(huì)員上傳分享,免費(fèi)在線閱讀,更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫(kù)。
1、摘要隨著計(jì)算機(jī)、多媒體等的迅速發(fā)展和應(yīng)用,數(shù)字字符的檢測(cè)和識(shí)別在當(dāng)今社會(huì)的許多領(lǐng)域都有著廣闊的應(yīng)用前景.盡管目前人們對(duì)數(shù)字字符的檢測(cè)和識(shí)別提出了許多方法,但涉及到具體問(wèn)題時(shí)仍有很多問(wèn)題存在,本文根據(jù)數(shù)字字符的特點(diǎn)做了以下幾方面工作.第一,對(duì)模式識(shí)別技術(shù)中的獨(dú)立分量分析算法(IndependentComponentAnalysis,ICA)進(jìn)行了研究,詳細(xì)地論述了ICA的原理及實(shí)現(xiàn)過(guò)程,系統(tǒng)地介紹了目前幾種主要ICA算法以及它們之間的內(nèi)在聯(lián)系,并重點(diǎn)分析了一種快速I(mǎi)CA實(shí)現(xiàn)算法FastICA,在此基礎(chǔ)上,利用一種改進(jìn)的FastICA算法進(jìn)行數(shù)字字符識(shí)別的應(yīng)
2、用研究,與一般的ICA算法相比,加快了收斂速度,提高了數(shù)字字符識(shí)別速度.第二,針對(duì)地形掃描圖的矢量化問(wèn)題,進(jìn)行了兩項(xiàng)預(yù)處理工作的研究.首先通過(guò)提取連通分量分離高程值標(biāo)記,其工作還包括地圖的預(yù)處理,如二值化等,二值化的效果影響后續(xù)工作的進(jìn)行,有效的二值化算法是高檢測(cè)率的保證;將經(jīng)過(guò)連通分量提取后的候選數(shù)字進(jìn)行膨脹,結(jié)構(gòu)元素選擇擬圓形,即消除了非數(shù)字字符要素,又可以計(jì)算出傾斜角度,為旋轉(zhuǎn)作好準(zhǔn)備.其次利用基于數(shù)值背景表達(dá)的方法對(duì)數(shù)字字符識(shí)別,與基于獨(dú)立分量分析的方法相比較,此方法計(jì)算較為簡(jiǎn)單.實(shí)驗(yàn)結(jié)果表明此方法具有很好的效果,對(duì)各種字體、各種字號(hào)的數(shù)字字符都
3、取得較好的識(shí)別效果.關(guān)鍵詞:數(shù)字字符檢測(cè);數(shù)字字符識(shí)別;獨(dú)立分量分析;數(shù)學(xué)形態(tài)學(xué);連通分量IAbstractWiththerapiddevelopmentandapplicationsofthecomputerandmultimedia,digitalcharactersdetectionandidentificationhavebroadapplicationinmanyareasintoday'ssociety.Althoughmanymethodshavebeenproposed,manyproblemsstillexistwhenthespeci
4、ficproblemsarerelated.Theworkaccordingtofeaturesofthedigitalcharactersinthispaperisasfollows:First,thispaperresearchesIndependentComponentAnalysis(ICA)algorithmofpatternrecognitiontechnologyandparticularlytreatsbasicprincipleandimplementationofICA,roundlyintroducescurrentseveralm
5、ainICAalgorithmsandtherelationofthem.analysisafastalgorithm,FastICA.Onthisbasis,thispaperresearchcharacterrecognitionapplicationusinganimprovedFastICAalgorithm.Itacceleratestheconvergenceofspeed,digitalcharacterrecognitionspeedthangeneralalgorithm.Second,Twopre-processingtechnolo
6、giesforvectorizingscannedtopographicmapsarepresented.Firstly,analgorithmforextractionofconnectedcomponentsbasedonmathematicalmorphologyisemployedtofindoutelevationvaluenotes;Thispaperalsoincludesmapspretreatment,suchasbinary.Binaryeffectsthefollowingwork,theeffectivebinaryalgorit
7、hmguaranteesahighdetectionrate.Afterextractingconnectedcomponent,dilatecandidatenumber,andstructuralelementschoosesrotundity.Itnotonlyeliminatenon-digitcharacterelementsbutalsocalculateangleofrotation.Secondly,recognizedigitalcharacterbasedonthenumericalexpressionofthebackground.
8、Compar-edthemethodofbasedontheindependen