利用人工神经网络应用相关度提高图像识别决策质量的新方法(IJIGSP-V6-N11-4)

I.J.Image,GraphicsandSignalProcessing,2014,11,25-35PublishedOnlineOctober2014inMECS()DOI:10.5815/ijigsp.2014.11.04Copyright©2014MECSI.J.Image,GraphicsandSignalProcessing,2014,11,25-35ANovelApproachforImageRecognitiontoEnhancetheQualityofDecisionMakingbyApplyingDegreeofCorrelationUsingArtificialNeuralNetworksRajuDaraResearchScholar,DepartmentofComputerScienceandEngineeringJawaharlalNehruTechnologicalUniversityKakinada,AndhraPradesh,[e-mail]olar,DepartmentofComputerScienceandEngineeringJawaharlalNehruTechnologicalUniversityKakinada,AndhraPradesh,[e-mail],DepartmentofComputerScienceandEngineeringJawaharlalNehruTechnologicalUniversity,Hyderabad.AndhraPradesh,[e-mail]—Manydiversifiedapplicationsdoexistinscience&technology,whichmakeuseoftheprimarytheoryofarecognitionphenomenonasoneofitssolutions.Recognitionscenarioisincorporatedwithasetofdecisionsandtheactionaccordingtothedecisionpurelyreliesonthequalityofextractedinformationonutmostapplications.Thus,thequalitydecisionmakingabsolutelyreckonsonprocessingmomentumandprecisionwhichareentirelycoupledwithrecognitionmethodology.Inthisarticle,alatestruleisformulatedbasedonthedegreeofcorrelationtocharacterizethegeneralizedrecognitionconstraintandtheapplicationisexploredwithrespecttoimagebasedinformationextraction.MachinelearningbasedperceptioncalledfeedforwardarchitectureofArtificialNeuralNetworkhasbeenappliedtoattaintheexpectedeminenceofelucidation.Theproposedmethodfurnishesextraordinaryadvantagessuchaslessmemoryrequirements,extremelyhighlevelsecurityforstoringdata,exceptionalspeedandgentleimplementationapproach.IndexTerms—Imagerecognition,informationretrieval,artificialneuralnetwork,degreeofcorrelation,fault-tolerance.I.INTRODUCTIONWiththeadvancementofdigitaltechnologyinthepresentlifewithoutinvolvementoftheimageisvirtuallynotpossible.Thisisoftenasaresultof,imagecarriescleardatainmostofthecasesnotsolelythatourbrainhasalotofstablestructuretoinvestigateandperceivetheimagefromapsychologicalpointofobservation.Foroperationalperspective,recognitionofanimageisextremelydifficultandcomputationallyexpansivedownside.Thechallengesofferedbyrecognitionmethodareaccuracy,operationalspeed,storagedemandandqualityinimplementation.Withrecognitionindifferentfields,variedapplicationsareintegratedlikeinhealthcare,automotive,business,internetbasedapplicationsandadvertisements,etc.Inthispaperatechniqueisproposedtorecognizetheimagefromadatabasewhereveritisstoredandinresultvariedactivitieswilloutlineaccordingtorequirementslikejustincaseofimagebaseddataretrievalorinindustrialautomation.Proposedmethodologyhasextensionofcapabilityofrecognitionforimagewithitsoutliervariationsandwithitspartialdata.Developedapproachhasgivenattentiontodeliverthegreatperformanceevenwithfaultsareofferedwithsolution.Thissortofqualityisenormouslycrucialforhardwarecharacterizationwhereverfaultsareinherentcharacteristicswiththemoment.A.BackgroundAltogethertheimagerecognitiontechniquesplannedandrequirepre-processingofimage,featureextractionstageandartificialneuralnetworkforclassificationpurpose[1].Pre-processingisanimportantstepthatenhancesthequalityandproducesanimageinwhichminutiaecanbedetectedcorrectly[2].Adatacleaninggraphwithdataqualityconstraintsisusedtohelpusersinidentifyingthepointsofthegraph,andtherecordsneedtheattentiontowardsmanualdatarepairstorepresentthe26ANovelApproachforImageRecognitiontoEnhancetheQualityofDecisionMakingbyApplyingDegreeofCorrelationUsingArtificialNeuralNetworksCopyright©2014MECSI.J.Image,GraphicsandSignalProcessing,2014,11,25-35requiredfeedbacktocleandataitemsmanually[3].Inthispaper,threemajordataminingmethods,namelyfunctionaldependencymining,associationruleminingandBaggingSVMsfordatacleaningarediscussed[4].Inthisarticle,thesystemusestheextract,transformandloadmethodsasthemainprocessmodel,apartfromthese,thetechniquesappliedtoscrubthedataare:parsingtechniquetoidentifythedirtydata,theregularexpressionmethodformatchingattributes,andk-NearestNeighboralgorithm[5].Organizationaldatabasesarecomprehendedwithdataofpoorquality,thispaperdevelopsaframeworkthatconsistsofsevenelements:managementresponsibilities,operationandassurancecosts,researchanddevelopment,production,distribution,personnelmanagement,andlegalfunctiontoanalyzedataquality[6].Informationquality(IQ)hasbecomeacriticalconcernoforganizationsandanactiveareaofManagementInformationSystems(MIS)research,themethodologywhichisusedherecalledasAIMQ(AIMQuality)toformabasisforIQassessmentandbenchmarking[7].Thisarticlepresentsamethodologytotestitsefficacythrougharigorouscasestudyandcontributestwokeyfeaturessuchas,firstoneisdevelopinga2X2conceptualmodelfordescribingIQwhichreferstothismodelastheproduct&serviceperformancemodelforinformationquality(PSP/IQ),andnextisintegratingtheIQdimension.PSP/IQmodelprovidesawaytocompareinformationqualityacrossorganizations,andtodevelopIQbenchmarks[8].Aninnovativetechniqueissupposedto