一种融合两种主成分分析的货币识别方法.pdfVIP

1702009，45(23) and ComputerE，谚neeringApplications计算机工程与应用 一种融合两种主成分分析的货币识别方法 贺建飚，崔艳丽 HE Yan—li Jian-biao．CUI 中南大学信息科学与工程学院，长沙410083 ofInformationScienceand South 410083，China College Engineering，CentralUniversity，Changsha E-mail：cuiyanli0733@126．eoln liE basedOnfusiontwodifferent Jian-biao，CUIYan-li．Currencyrecognition component principal and EngineeringApplications。2009。45(23)：170-172． Abstract：Inordertoextractthemostusefulfeaturesof novelmethodof is of recognition currency，a currency proposed．First kindsof usedto reducethedimensionalitiesofthe all，two ComponentAnalysis(PCA)are originalimage Principle respectively vector setattribute is introducedto the offeaturevectors．At optimization optimize last，canonical space．Then，FOUgh parameters correlationis tofusethesefeatures．TheresultsshOWthatthe ofthe methodcallextract analysisadopted performanceproposed themoreusefulfeatures，andthe resultsarebetterthanthoseof onekindofPCA．Inaddition，whenthenumber recognition using ofthe setis20．the rateis98．78％． training recognition set；canonicalcorrelation fusion Keywords：currencyrecognition；PrincipleComponentAnalysis(PCA)；rough analysis；feature 摘要：为了充分提取货币图像的有效特征，提出了一种新颖的货币识别算法。首先，利用两种主成分分析分剐降低原始图像空间 的维数；然后，对所得特征参数属性用粗糙集决策表约简原理进行优化；最后，运用典型相关分析原理融合特征级数据，并给出最 终识别效果。实验结果表明，该算法在充分提取货币图像特征的基础上，能很好地融合其特