外文文獻翻譯成品：基于人臉識別的移動自動課堂考勤管理系統(tǒng)(中英文雙語對照)

外文標題：FaceRecognition-BasedMobileAutomaticClassroomAttendanceManagementSystem外文作者：RefikSamet,MuhammedTanriverdi文獻出處:2018InternationalConferenceonCyberworlds(如覺得年份太老，可改為近2年，畢竟很多畢業(yè)生都這樣做)英文2937單詞，20013字符(字符就是印刷符)，中文4819漢字。FaceRecognition-BasedMobileAutomaticClassroomAttendanceManagementSystemAbstract—Classroomattendancecheckisacontributingfactortostudentparticipationandthefinalsuccessinthecourses.Takingattendancebycallingoutnamesorpassingaroundanattendancesheetarebothtime-consuming,andespeciallythelatterisopentoeasyfraud.Asanalternative,RFID,wireless,fingerprint,andirisandfacerecognition-basedmethodshavebeentestedanddevelopedforthispurpose.Althoughthesemethodshavesomepros,highsysteminstallationcostsarethemaindisadvantage.Thepresentpaperaimstoproposeafacerecognition-basedmobileautomaticclassroomattendancemanagementsystemneedingnoextraequipment.Tothisend,afilteringsystembasedonEuclideandistancescalculatedbythreefacerecognitiontechniques,namelyEigenfaces,FisherfacesandLocalBinaryPattern,hasbeendevelopedforfacerecognition.Theproposedsystemincludesthreedifferentmobileapplicationsforteachers,students,andparentstobeinstalledontheirsmartphonestomanageandperformthereal-timeattendance-takingprocess.TheproposedsystemwastestedamongstudentsatAnkaraUniversity,andtheresultsobtainedwereverysatisfactory.Keywords—facedetection,facerecognition,eigenfaces,fisherfaces,localbinarypattern,attendancemanagementsystem,mobileapplication,accuracyI.INTRODUCTIONMosteducationalinstitutionsareconcernedwithparticipationincoursessincestudentparticipationintheclassroomleadstoeffectivelearningandincreasessuccessrates[1].Also,ahighparticipationrateintheclassroomisamotivatingfactorforteachersandcontributestoasuitableenvironmentformorewillingandinformativeteaching[2].Themostcommonpracticeknowntoincreaseattendanceinacourseistakingattendanceregularly.Therearetwocommonwaystocreateattendancedata.Someteachersprefertocallnamesandputmarksforabsenceorpresence.Otherteachersprefertopassaroundapapersigningsheet.Aftergatheringtheattendancedataviaeitherofthesetwomethods,teachersmanuallyenterthedataintotheexistingsystem.However,thosenon-technologicalmethodsarenotefficientwayssincetheyaretime-consumingandpronetomistakes/fraud.Thepresentpaperaimstoproposeanattendance-takingprocessviatheexistingtechnologicalinfrastructurewithsomeimprovements.Afacerecognition-basedmobileautomaticclassroomattendancemanagementsystemhasbeenproposedwithafacerecognitioninfrastructureallowingtheuseofsmartmobiledevices.Inthisscope,afilteringsystembasedonEuclideandistancescalculatedbythreefacerecognitiontechniques,namelyEigenfaces,Fisherfaces,andLocalBinaryPattern(LBP),hasbeendevelopedforfacerecognition.Theproposedsystemincludesthreedifferentapplicationsforteachers,students,andparentstobeinstalledontheirsmartphonestomanageandperformareal-timepollingprocess,datatracking,andreporting.Thedataisstoredinacloudserverandaccessiblefromeverywhereatanytime.Webservicesareapopularwayofcommunicationforonlinesystems,andRESTfulisanoptimalexampleofwebservicesformobileonlinesystems[3].Intheproposedsystem,RESTfulwebserviceswereusedforcommunicationamongteacher,student,andparentapplicationsandthecloudserver.Attendanceresultsarestoredinadatabaseandaccessiblebytheteacher,studentandparentmobileapplications.Thepaperisorganisedasfollows.SectionIIprovidesabriefliteraturesurvey.SectionIIIintroducestheproposedsystem,andsectionIVfollowsbyimplementationandresults.Thelastsectiongivesthemainconclusions.LITERATURESURVEYFingerprintreadingsystemshavehighinstallationcosts.Furthermore,onlyonestudentatatimecanuseaportablefingerrecognitiondevice,whichmakesitatime-consumingprocess[4].Inthecaseofafixedfingerrecognitiondeviceattheentranceoftheclassroom,attendance-takingshouldbedoneundertheteacher'ssupervisionsothatstudentsdonotleaveafterthefingerrecognition,whichmakestheprocesstime-consumingforboththeteacherandthestudents.IncaseofRFIDcardreadingsystems,attendance-takingisavailableviathecardsdistributedtostudents[5].Insuchsystems,studentsmayresorttofraudulentmethodsbyreadingtheirfriends'cards.Also,ifastudentforgetshis/hercard,anon-trueabsencemaybesavedinthesystem.ThedisadvantageoftheclassroomscanningsystemswithBluetoothorbeaconmethodsisthateachstudentmustcarryadevice.BecausethefieldlimitoftheBluetoothLowEnergy(BLE)systemcannotbedetermined,studentswhoarenotintheclassroomatthemomentbutarewithintheBluetootharealimitsmayappeartobepresentintheattendancesystem[6].Therearedifferentmethodsofclassroomattendancemonitoringusingfacerecognitiontechnology.Oneoftheseisacameraplacedattheclassroomentranceandthestudentsenteringtheclassroomareregisteredintothesystembyfacerecognition[7].However,inthissystemstudents’facescouldberecognised,althoughstudentscanleavetheclassroomafterwards,anderrorscanoccurinthepollinginformation.Anothermethodistheobservationcarriedoutwithacameraplacedintheclassroomandtheclassroomimagetakenduringthecourse.Inthiscase,thecamerasusedinthesystemneedtobechangedfrequentlytokeepproducingbetterqualityimages.Therefore,thissystemisnotveryusefulandcanbecomecostly.Inadditiontoalltheaforementioneddisadvantages,themostcommondisadvantageisthatallthesemethodsneedextraequipment.Theproposedsystemhasbeendevelopedtoaddressthesedisadvantages.Themainadvantagesoftheproposedsystemareflexibleusage,noequipmentcosts,nowastedtime,andeasyaccessibility.PROPOSEDSYSTEMArchitectureoftheProposedSystemTheproposedsystem'sarchitecturebasedonmobilityandflexibilityisshowninFig.1.Figure1.SystemArchitectureThesystemconsistsofthreelayers:ApplicationLayer,CommunicationLayer,andServerLayer.ApplicationLayer:Intheapplicationlayer,therearethreemobileapplicationsconnectedtothecloudserverbywebservices.a)TeacherApplication:Theteacheristheheadofthesystem,sohe/shehastheprivilegetoaccessallthedata.Byhis/hersmartmobiledevice,he/shecantakeaphotoofstudentsinaclassroomatanytime.Afterthetakingthephotograph,theteachercanusethisphototoregisterattendance.Forthisaim,thephotoissenttothecloudserverforfacedetectionandrecognitionprocessing.Theresultsaresavedintoadatabasetogetherwithallthereachabledata.Theteachergetsaresponsebythemobileapplicationandcanimmediatelyseetheresults.Theteachercanalsocreateastudentprofile,addaphotoofeachstudent,andaddorremoveastudentto/fromtheirclassrosters.He/shecanaswellcreateanddeletecourses.Eachcoursehasauniquesix-charactercode.Theteachercansharethiscodewithhis/herstudentssotheycanaccesstheirattendanceresultsviathestudentapplication.Theteachercanaccesstoalldataandresultsbasedoneachstudent'srecognizedphotostampedwithadate.Additionally,anemailmessagewithattendancedataofaclassinExcelformatcanberequested,whiletheanalyticsoftheattendanceresultsisprovidedintheapplication.b)StudentApplication:Studentscansignincourseswiththeteacher'semailaddressandthesix-charactercoursecode.Theycanaddtheirphotosbytakingaphotoora3-secondlongvideo.Incaseoferrors,theiruploadedphotoscanbedeleted.Studentscanonlyseelimitedresultsoftheattendance-takingprocessrelatedtotheirattendance.Toprotectpersonalprivacy,theclassphotosanddetectedportraitphotosofeachstudentcanbeaccessedonlybytheteacher.Ifstudentsarenotintheclassroomwhenanattendance-checkisperformed,theyarenotifiedoftheattendance-check.Incaseoferrors(ifastudentispresent,butnotdetectedbythesystem),he/shecannotifytheteachersohe/shecanfixtheproblem.c)FamilyApplication:Parentscanseetheirchildren'sattendanceresultsforeachclass.Additionalchildrenprofilescanbeaddedintothesystem.Eachparentisaddedtothestudent'sapplicationwithname,surname,andemailaddress.Whenastudentaddshis/herparents,theyareautomaticallyabletoseetheattendanceresults.Theyarealsonotifiedwhentheirchildisnotintheclassroom.CommunicationLayer:RESTfulwebservicesareusedtocommunicatebetweentheapplicationsandserverlayers.RequestsaresentbythePOSTmethod.EachrequestissentwithauniqueIDoftheauthoriseduserofthesession.Onlytheauthoriseduserscanaccessandrespondthethedatatowhichtheyhaverighttoaccess.Duetoitsflexibilityandfastperformance,JSONisusedasthedataformatforwebservicesresponse[8].Withthisabstractwebservicelayer,thesystemcaneasilybeusedforanewitemintheapplicationlayer,suchaswebpagesoranewmobileoperatingsystem.ServerLayer:Theserverlayerisresponsibleforhandlingtherequestsandsendingtheresultstotheclient.Facedetectionandrecognitionalgorithmsareperformedinthislayerandmorethan30differentwebservicesarecreatedforhandlingdifferentrequestsfrommobileapplications.FaceDetectionAccurateandefficientfacedetectionalgorithmsimprovetheaccuracylevelofthefacerecognitionsystems.Ifafaceisnotdetectedcorrectly,thesystemwillfailitsoperation,stopprocessing,andrestart.Knowledge-based,feature-based,template-based,andstatistics-basedmethodsareusedforfacedetection[9].Sincetheclassroomphotoistakenundertheteacher'scontrol,posevariationscouldbelimitedtoasmallrange.Viola-JonesfacedetectionmethodwithAda-boosttrainingisshownasthebestchoiceforreal-timeclassattendancesystems[9,10].Inthemostbasicsense,thedesiredobjectsarefirstlyfoundandintroducedaccordingtoacertainalgorithm.Afterwards,theyarescannedtofindmatcheswithsimilarshapes[11].FaceRecognitionTherearetwobasicclassificationsoffacerecognitionbasedonimageintensity:feature-basedandappearance-based[12].Feature-basedapproachestrytorepresent(approximate)theobjectascompilationsofdifferentfeatures,forexample,eyes,nose,chin,etc.Incontrast,theappearance-basedmodelsonlyusetheappearancecapturedbydifferenttwo-dimensionalviewsoftheobject-of-interest.Feature-basedtechniquesaremoretime-consumingthanappearance-basedtechniques.Thereal-timeattendancemanagementsystemrequireslowcomputationalprocesstime.Therefore,threeappearance-basedfacerecognitiontechniquessuchasEigenfaces,FisherfacesandLBPareusedinthetestedsystem.Fisherfacesandeigenfacestechniqueshaveavaryingsuccessrate,dependingondifferentchallenges,likeposevariation,illumination,orfacialexpression[13].Accordingtoseveralpreviousstudies,facerecognitionusingLBPmethodgivesverygoodresultsregardingspeedanddiscriminationperformanceaswellasindifferentlightingconditions[14,15].Euclideandistanceiscalculatedbyfindingsimilaritiesbetweenimagesforfacerecognition.AfilteringsystembasedonEuclideandistancescalculatedbyEigenfaces,FisherfacesandLBPhasbeendevelopedforfacerecognition.Accordingtothed,,mnsf,sandEigenfacesalgorithmsareevaluatedindefinedorder.IftheEuclideandistanceofLBPalgorithmislessthan40;elseifEuclideandistanceofFisherfacesalgorithmislessthan250;elseifEuclideandistanceofEigenfacesalgorithmislessthan1500,recognizedfaceisrecordedastherightmatch.Secondly,ifthecalculatedEuclideandistancesbythethreemethodsaregreaterthantheminimumEuclideandistances,thesecondlevelEuclideandistances(40-50(forLBP),250-400(forFisherfaces),1500-1800(forEigenfaces))areevaluatedinthesameway.Ifthesecondlevelconditionsarealsonotmet,thefilterreturnsthewrongmatch.Thirdly,ifanytwoalgorithmsgivethesamematchresult,thematchisrecordedcorrectly.Finally,ifnoconditionsaremet,thepriorityisgiventotheLBPalgorithmandthematchisrecordedcorrectly.escarchitectureaimedr,,andtrequiringnoextraequipment.Atthesametime,itsobjectivewastoprovideaccesstoallusersatanytime.Thesystemthusoffersareal-timeattendancemanagementsystemtoallitsusers.IV.IMPLEMENTATIONANDRESULTSThefollowingplatformwasused.Thecloudserverhasa2.5GHzwith4-coreCPU,8GBRAM,and64-bitoperatingsystemcapacity.Viola-JonesfacedetectionalgorithmandEigenfaces,FisherfacesandLBPfacerecognitionalgorithmswereimplementedbasedonOpenCV.TestsweredonewithbothiOSandANDROID.Fortydifferentattendancemonitoringtestswereperformedinarealclassroom,including11students,and264students’facesweredetected.TablesI,II,andIIIshowdetectionandrecognitionaccuracyofallthreedifferenttypesoftestedalgorithmsrelatedtotheEuclideandistance.Priorityorderingfor3algorithmswasarrangedaccordingtoaccuracyrateforeachinterval.Intestresults,123,89,and85falserecognitionsweredetectedforEigenfaces,Fisherfacesandrespectively.Bythehelpofthedevelopedfilteringsystem,thenumberoffalserecognitionsdecreasedto65.Outof40implementedattendancemonitoringtests,10wereconductedwith1facephotoofeachstudentindatabaseinStep-I,20wereconductedwhenthenumberoffacephotosincreasedupto3inStep-II,and10recognitionprocesseswereconductedwithmorethan3facephotosindatabaseinStep-III.TableIVshowstheobtainedresults.Themostimportantlimitationoftestedattendancemonitoringprocessisdecreasedsuccesswithincreasingdistancebetweenthecameraandstudents.Theresultsregardingstudentssittinginfrontseatsaremoreaccurateincomparisontoresultsregardingstudentssittingintheback.Secondly,theaccuracyratesmayhavedecreasedduetotheblurringcausedbyvibrationwhilethephotowastaken.Thirdly,insomecasesonepartofthestudent'sfacemaybecoveredbyanotherstudentsittinginfrontofhim/her,whichmayhamperasuccessfulfacerecognitionprocess.Sincetheclassroomphotosaretakeninuncontrolledenvironments,theilluminationandposecould,toalargeextent,affecttheaccuracyrate.Thedevelopedfilteringsystemminimizestheseeffects.Toincreaseaccuracy,posetolerantfacerecognitionapproachmayalsobeused[16,17].V.CONCLUSIONSThepresentpaperproposesaflexibleandreal-timefacerecognition-basedmobileattendancemanagementsystem.AfilteringsystembasedonEuclideandistancescalculatedbyEigenfaces,Fisherfaces,andLBPhasbeendeveloped.Theproposedsystemeliminatesthecostforextraequipment,minimizesattendance-takingtime,andallowsuserstoaccessthedataanytimeandanywhere.Smartdevicesareveryuser-friendlytoperformclassroomattendancemonitoring.Teachers,students,andparentscanusetheapplicationwithoutanyrestrictionsandinreal-time.Sincetheinternetconnectionspeedhasbeensteadilyincreasing,highquality,largerimagescanbesenttotheserver.Inaddition,processorcapacityoftheserversisalsoincreasingondailybasis.Withthesetechnologicaldevelopments,theaccuracyrateoftheproposedsystemwillalsobeincreased.Facerecognitioncouldbefurthertestedbyotherfacerecognitiontechniques,suchasSupportVectorMachine,HiddenMarkovModel,NeuralNetworks,etc.Additionally,detectionandrecognitionprocessescouldbeperformedonsmartdevicesoncetheirprocessorcapacityissufficientlyincreased.REFERENCES[1]L.Stanca,"TheEffectsofAttendanceonAcademicPerformance:PanelDataEvidenceforIntroductoryMicroeconomics,"J.Econ.Educ.,vol.37,no.3,pp.251–266,2006.PaniandKishore,"AbsenteeismandperformanceinaquantitativemoduleAquantileregressionanalysis,"JournalofAppliedResearchinHigherEducation,vol.8no.3,pp.376-389,2016.[3]U.Thakar,A.Tiwari,andS.Varma,"OnCompositionofSOAPBasedandRESTfulServices,"IEEE6thInt.ConferenceonAdvancedComputing(IACC),2016.BasheerandC.V.Raghu,"Fingerprintattendancesystemforclassroomneeds,"AnnualIEEEIndiaConference(INDICON),pp.433-438,2012.[5]S.Konatham,B.S.Chalasani,N.Kulkarni,andT.E.Taeib,―AttendancegeneratingsystemusingRFIDandGSM,‖IEEELongIslandSystems,ApplicationsandTechnologyConference(LISAT),2016.[6]S.Noguchi,M.Niibori,E.Zhou,andM.Kamada,"StudentAttendanceManagementSystemwithBluetoothLowEnergyBeaconandAndroidDevices,"18thInternationalConferenceonNetwork-BasedInformationSystems,pp.710-713,2015.[7]S.ChintalapatiandM.V.Raghunadh,―Automatedattendancemanagementsystembasedonfacerecognitionalgorithms,‖IEEEInt.ConferenceonComputationalIntelligenceandComputingResearch,2013.[8]G.Wang,"ImprovingDataTransmissioninWebApplicationsviatheTranslationbetweenXMLandJSON,"ThirdInt.ConferenceonCommunicationsandMobileComputing(CMC),pp.182-185,2011.[9]X.Zhu,D.Ren,Z.Jing,L.Yan,andS.Lei,"ComparativeResearchoftheCommonFaceDetectionMethods,"2ndInternationalConferenceonComputerScienceandNetworkTechnology,pp.1528-1533,2012.[10]V.GuptaandD.Sharma,―AStudyofVariousFaceDetectionMethods,‖InternationalJournalofAdvancedResearchinComputerandCommunicationEngineeringvol.3,no.5,pp.6694-6697,2014.[11]P.ViaolaandM.J.Jones,―RobustReal-TimeFaceDetection,‖InternationalJournalofComputerVision,vol.57,no.2,pp.137-154,2004.[12]L.Masupha,T.Zuva,S.Ngwira,andO.Esan,―Facerecognitiontechniques,theiradvantages,disadvantagesandperformanceevaluation,‖Int.ConferenceonComputing,CommunicationandSecurity(ICCCS),2015.[13]J.Li,S.Zhou,andC.Shekhar,"Acomparisonofsubspaceanalysisforfacerecognition,"InternationalConferenceonMultimediaandExpo,ICME'03,Proceedings,vol.3,pp.121-124,2003.[14]T.Ahonen,A.Hadid,andM.Pietikainen,―FacedescriptionwithLocalBinaryPatterns,‖IEEETransactionsonPatternAnalysisandMachineIntelligence,vol.28,no.12,pp.2037-2041,2006.[15]T.Ahonen,A.Hadid,M.Pietikainen,andT.Maenpaa,―Facerecognitionbasedontheappearanceoflocalregions,‖Proceedingsofthe17thInt.ConferenceonPatternRecognition,vol.3,pp.153-156,2004.[16]R.Samet,S.Sakhi,andK.B.Baskurt,―AnEfficientPoseTolerantFaceRecognitionApproach‖,TransactionsonComput.ScienceXXVI,LNCS9550,pp.161-172,2016.[17]R.Samet,G.S.Shokouh,J.Li,―ANovelPoseTolerantFaceRecognitionApproach‖,2014InternationalConferenceonCyberworlds,pp.308-312,2014.基于人臉識別的移動自動課堂考勤管理系統(tǒng)摘要-課堂出勤檢查是學生參與和課程最終成功的一個因素。通過呼叫姓名或繞過考勤表來參加考勤都是非常耗時的，尤其是后者容易欺詐。作為替代方案，已經為此目的測試和開發(fā)了RFID，無線，指紋，虹膜和基于面部識別的方法。雖然這些方法有一些優(yōu)點，但高系統(tǒng)安裝成本是主要缺點。本文旨在提出一種基于人臉識別的移動自動教室考勤管理系統(tǒng)，無需額外設備。為此，已經開發(fā)了一種基于歐幾里德距離的濾波系統(tǒng)，該系統(tǒng)通過三種人臉識別技術（即特征臉，Fisherfaces和局部二值模式）計算，用于人臉識別。建議的系統(tǒng)包括三個不同的移動應用程序，供教師，學生和家長安裝在他們的智能手機上，以管理和執(zhí)行實時考勤過程。擬議的系統(tǒng)在安卡拉大學的學生中進行了測試，結果非常令人滿意。關鍵詞-人臉檢測，人臉識別，特征臉，漁民，局部二值模式，考勤管理系統(tǒng)，移動應用，準確性一、引言大多數教育機構都關注學生參與課程，因為學生參與課堂會導致有效學習并提高成[1]。此外，課堂上的高參與率是教師的激勵因素，并為更加自愿和信息豐富的教學創(chuàng)造了合適的環(huán)境[2]。已知增加課程出勤率的最常見做法是定期參加。創(chuàng)建出勤數據有兩種常用方法。有些老師喜歡打電話給姓名并留下缺席或存在的標記。其他老師喜歡傳遞紙質簽名表。在通過這兩種方法之一收集出勤數據后，教師手動將數據輸入現有系統(tǒng)。然而，這些非技術方法并不是有效的方法，因為它們耗時并且容易出錯或欺詐。本文旨在通過現有的技術基礎設施提出一個考勤接受過程，并進行一些改進。已經提出了一種基于面部識別的移動自動教室考勤管理（即特征臉，Fisherfaces和局部二值模式（LBP）計算，用于人臉識別。建議的系統(tǒng)包括三個不同的應用程序，供教師，學生和家長安裝在他們的智能手機上，以管理和執(zhí)行實時輪詢過程，數據跟蹤和報告。數據存儲在云服務器中，可隨時隨地訪問。Web服務是在線系統(tǒng)的一種流行的通信方式，RESTful是移動在線系統(tǒng)的Web服務的最佳示例[3]。在所提出的系統(tǒng)中，RESTfulWeb服務用于教師，學生和父應用程序與云服務器之間的通信。出勤結果存儲在數據庫中，可供教師，學生和家長移動應用程序訪問。該論文的組織如下。第二節(jié)提供了簡要的文獻調查。第三節(jié)介紹了擬議的系統(tǒng)，第四節(jié)介紹了實施和結果。最后一節(jié)給出了主要結論。二、文獻綜述指紋讀取系統(tǒng)具有高安裝成本。此外，一次只有一個學生可以使用便攜式手指識別設備，這使得這是一個耗時的過程[4]。在教室入口處固定手指識別裝置的情況下，應在老師的監(jiān)督下進行考勤，以便學生在手指識別后不會離開，這使得教師和教師都需要花費大量時間。學生們。在RFID卡讀取系統(tǒng)的情況下，通過分發(fā)給學生的卡可以獲得考勤[5]。在這樣的系統(tǒng)中，學生可以通過閱讀朋友的卡片來采用欺詐手段。此外，如果學生忘記了他/她的卡，可以在系統(tǒng)中保存非真實缺席。具有藍牙或信標方法的教室掃描系統(tǒng)的缺點是每個學生必須攜帶設備。由于無法確定藍牙低功耗（BLE）系統(tǒng)的字段限制，目前不在教室但在藍牙區(qū)域范圍內的學生可能會出現在考勤系統(tǒng)中[6]。使用人臉識別技術有不同的課堂考勤監(jiān)控方法。其中一個是放置在教室入口處的攝像機，進入教室的學生通過人臉識別登記到系統(tǒng)中[7]。然而，在這個系統(tǒng)中，學生的面孔可以被識別，盡管學生之后可以離開教室，并且輪詢信息中可能會出現錯誤。另一種方法是使用放置在教室中的攝像機和在課程期間拍攝的教室圖像進行觀察。在這種情況下，系統(tǒng)中使用的攝像機需要經常更換，以保持產生更高質量的圖像。因此，該系統(tǒng)不是非常有用并且可能變得昂貴。除了上述所有缺點之外，最常見的缺點是所有這些方法都需要額外的設備。已經開發(fā)出所提出的系統(tǒng)以解決這些缺點。該系統(tǒng)的主要優(yōu)點是使用靈活，無需設備成本，不浪費時間，易于訪問。三、擬議的系統(tǒng)擬議系統(tǒng)的架構基于移動性和靈活性的所提出的系統(tǒng)架構如圖1所示。.圖1.系統(tǒng)架構該系統(tǒng)由三層組成：應用層，通信層和服務器層。應用層：在應用層，有三個移動應用通過Web服務連接到云服務器。a）教師申請：教師是系統(tǒng)的負責人，因此他/她有權訪問所有數據。通過他/她的智能移動設備，他/她可以隨時在教室中拍攝學生的照片。拍照后，老師可以用這張照片登記出勤。為此目的，將照片發(fā)送到云服務器以進行面部檢測和識別處理。結果與所有可到達的數據一起保存到數據庫中。教師通過移動應用程序獲得響應，并可立即查看結果。教師還可以創(chuàng)建學生檔案，添加每個學生的照片，以及在班級名單中添加或刪除學生。他/她也可以創(chuàng)建和刪除課程。每門課程都有一個獨特的六字符代碼。教師可以與他她的學生分享此代碼，以便他們可以通過學生應用程序訪問他們的出勤結果。教師可以根據每個學生認可的帶有日期標記的照片訪問所有數據和結果。另外，可以請求具有Excel格式的班級的出勤數據的電子郵件消息，同時在應用程序中提供出勤結果的分析。b）學生申請：學生可以使用教3秒長的視頻來添加照片。如果出現錯誤，可以刪除他們上傳的照片。學生只能看片和檢測到的肖像照片只能由老師訪問。如果學生在進行出勤檢查時不在教室，則會通知他們出勤檢查。如果出現錯誤（如果學生在場，但系統(tǒng)未檢測到他/她可以通知教師，以便他她可以解決問題。家庭申請：家長可以看到孩子每節(jié)課的出勤率?？梢詫⑵渌优渲梦募砑拥较到y(tǒng)中。每個父母都會添加到她的父母時，他們會自動查看出勤結果。當他們的孩子不在教室時，他們也會收到通知。3）通信層：RESTfulWeb服務用于在應用程序和服務器層之間進行通信。請求由POST方法發(fā)送。每個請求都使用會話的授權用戶的唯一ID發(fā)送。只有授權用戶才能訪問和響應他們有權訪問的數據。由于其靈活性和快速性能，JSON被用作Web服務響應的數據格式[8]。利用這種抽象的Web服務層，系統(tǒng)可以很容易地用于應用層中的新項目，例如網頁或新的移動操作系統(tǒng)。3）部檢測和識別算法，并且創(chuàng)建了30多種不同的Web服務，用于處理來自移動應用程序的不同請求。人臉檢測準確有效的人臉檢測算法提高了人臉識別系統(tǒng)的準確度。如果未正確檢測到面部，[9]。由于教室照片是在教師的控制下拍攝的，因此姿勢變化可能僅限于小范圍。使用Ada-boost訓練的Viola-Jones面部檢測方法被視為實時課堂考勤系統(tǒng)的最佳選擇[9,10]。在最基本的意義上，首先根據某種算法找到并引入所需的對象。然后，他們被掃描以找到具有相似形狀的匹配[11].人臉識別[12]?；谔卣鞯姆椒ㄔ噲D將對象表示（近似）基于特征的技術比基于外觀的技術更耗時。實時考勤管理系統(tǒng)需要較低的計算處理時間。因此，在測試系統(tǒng)中使用三種基于外觀的面部識別技術，例如Eigenfaces，Fisherfaces和LBP。Fisherfaces和eigenfaces技術具有不同[13]。根據之前的幾項研究，使用LBP方法進行人臉識別可以在速度和辨別性能以及不同的光照條件[14,15]。通過尋找用于面部識別的圖像之間的相似性來計算歐幾里德距離?；跉W幾里德距離的特征面，Fisherfaces和LBP計算的濾LBP，Fisherfaces和Eigenfaces算法的最小歐幾里德距離。如果LBP算法的歐幾里德距離小于40;否則，如果Fisherfaces算法的歐幾里德距離小于250;否則，如果Eigenfaces算法的歐幾里德距離小于1500，則識別的面部被記錄為正確的第二級歐幾里德距離（對于，（對于，（對于特征臉）是以同樣的方式評估。如果也未滿足第二級條件，則過濾器返回錯誤的匹配。第三，如果任何兩個算法給出相同的匹配結果，則正確記錄匹配。最后，如果沒有滿足條件，則優(yōu)先考慮LBP架構旨在通過不需要額外設備來實現靈活性，移動性和低成本。同時，其目標是隨時為所有用戶提供訪問權限。因此，該系統(tǒng)為所有用戶提供實時考勤管理系統(tǒng)。四、實施和結果使用以下平臺。云服務器具有2.5GHz，4核CPU，8GBRAM和64位操作系統(tǒng)容量。基于OpenCV實現了Viola-Jones人臉檢測算法和特征臉，Fisherfaces和LBP人臉識別算法。iOS和ANDROID都進行了測試。在真實的教室中進行了40次不同的出勤監(jiān)測測試，包括11名學生，并檢測到264名學生的面部。表II和III顯示了與歐氏距離相關的所有三種不同類型的測試算法的檢測和識別準確度。根據每個間隔的準確率安排3種算法的優(yōu)先級排序。在測試結果中，分別檢測到特征臉，Fisherfaces和LBP的123,89和85個錯誤識別。在開發(fā)的過濾系統(tǒng)的幫助下，錯誤識別的數量減少到65個。在40個實施的考勤監(jiān)測測試中，10個進行了步驟I1張面部照片，當數量為20時進行在步驟II中，面部照片增加到3，并且在步驟III中在數據庫中使用超過3張面部10IV顯示了獲得的結果。測試考勤監(jiān)控過程中最重要的限制是隨著攝像機與學生之間距離的增加而降低成功率。關于坐在前排座位上的學生的結果與坐在后排的學生的結果相比更準確。其次，由于在拍攝照片時由振動引起的模糊，準確率可能已經降低。第三，在某些情況下，學生面部的一部分可能被坐在他/她面前的另一名學生所覆蓋，這可能妨礙成功的面部識別過程。由于教室照片是在不受控制的環(huán)境中拍攝的，因此照明和姿勢可能在很大程度上影響準確率。開發(fā)的過濾系統(tǒng)使這些影響最小化。為了提高準確性，也可以使用姿勢容忍的人臉識別方法[16,17]。五、結論本文提出了一種靈活，實時的基于人臉識別的移動考勤管理系統(tǒng)。已經開發(fā)了基于歐幾里德距離的過濾系統(tǒng)，該過濾系統(tǒng)由特征臉，Fisherfaces和LBP計算。所提出的系統(tǒng)消除了額外設備的成本，最大限度地減少了考勤時間，并允許用戶隨時隨地訪問數據。智能設備非常易于用戶進行課堂出勤監(jiān)控。教師，學生和家長可以毫無限制地實時使用該應用程序。由于互聯網連接速度一直在穩(wěn)步增長，因此可以向服務器發(fā)送高質量，大圖像。此外，服務器的處理器容量也在日益增加。通過這些技術發(fā)展，所提出的系統(tǒng)的準確率也將提高。人臉識別可以通過其他人臉識別技術進一步測試，例如支持向量機，隱馬爾可夫模型，神經網絡等。另外，一旦處理器容量充分增加，就可以在智能設備上執(zhí)行檢測和識別過程。參考文獻[1]L.Stanca,"TheEffectsofAttendanceonAcademicPerformance:PanelDataEvidenceforIntroductoryMicroeconomics,"J.Econ.Educ.,vol.37,no.3,pp.266,2006.PaniandKishore,"AbsenteeismandperformanceinaquantitativemoduleAquantileregressionanalysis,"JournalofAppliedResearchinHigherEducation,vol.8no.3,pp.376-389,2016.[3]U.Thakar,A.Tiwari,andS.Varm