Computer vision Matlabpresentation116.docx

Computer vision Matlabpresentation116.docx

《Computer vision Matlabpresentation116.docx》由会员分享，可在线阅读，更多相关《Computer vision Matlabpresentation116.docx（9页珍藏版）》请在冰豆网上搜索。

ComputervisionMatlabpresentation116

Matlabpresentation

ComputerVision

220111XX刘XX

2014/10/27

TodayIwilldoapresentationaboutComputerVision.Thisistheoutlineofmypresentation.First,IwilldoabriefintroductionaboutComputerVisionanditsapplications.ThenIwillshowyousomethingaboutOpenCV.Finally,wewilltalksomethingaboutthedevelopmentofComputerVision.

NowIwilldoabriefintroductionaboutComputerVision.So,whatisComputervision?

TherearesomepicturesabouttheapplicationsofComputerVision,forexample,robotics[rəʊ'bɒtɪks]机器人学andverifyfaces.

Part1AbriefintroductionaboutComputerVision

ComputerVision,alsoknownas‘MachineVision’,isascienceaboutstudyinghowtomakeamachineto‘see’.Furthermore,thatis,insteadofthehumaneyes,wecanusecamerasandcomputerstotrack,measureandidentifythetargetandfurthertodographics['græfɪks]withcomputerprocessingtomakethetargetmoresuitableforthehumaneyetodetectortransmitimagestotheinstrument.

Asascientificdiscipline,computervisionresearchesrelevanttheoriesandtechniquesandtriestobuildanartificialintelligencesystemswhichcanget'information'fromtheimageormultidimensionalimagedata.Asfortheperception [pə'sepʃ（ə）n]感知canbeseenasasubjectofextractinginformationfromthesenses,sothecomputervisioncanbeseenasascienceofhowtomakeamanual['mænjʊ（ə）l]systemto‘sense’fromimagesormultidimensionalimagedata.

Allinall,thepurposeofComputerVisionistomimic ['mɪmɪk]模仿thefunctionsofhumanvisionsystemasmoreaspossible.AndsotheguidingprincipleofComputerVisionisthat‘Iftheeyecandoit,soshouldthemachine’.

Finally,letustalkaboutthekeycomponentsofComputerVision.ComputerVisionconsistsofDigitalCamera,Pre-procesing，FeatureExtractionandObjectRecognition.Furthermore,theycanalsobeclassifiedasLow-LevelVision,Mid-LevelVisionandHigh-LevelVision.

Thestructureandalgorithms ['ælgərɪð（ə）mz]算法donotnecessarilymimictheactualprocessofhumanvisualperception,butmayutilisesomeresultsofbiologicalandcognitive ['kɒɡnɪtɪv]认知的studies.

Nextstage,letustalkabouttheapplicationsofComputerVision.

Part2ApplicationsofComputerVision

TherearemanyapplicationsofComputerVision.Suchas,robotics,videosurveillance[sə'veɪl（ə）ns;-'veɪəns]监视,intelligentcruisecontrol智能巡航体系,automaticparking,automaticassemblylines自动装配生产线,automaticvisualinspection,targetdetection,human-computerinterfaceandsoon.

Nextstage,Iwillgiveyousomeexamplesofitsapplications.First,letuspayattentiontoAutomaticVisualInspection自动视觉检测orverifyfaces.Justasthepicturesays,whenweopenorloginthecomputer,thecomputercanrecognizeanddetectthefaceofusers.Andthecomputerjustuseourfacetobeapassword,sodifferentusershavedifferentaccountscorrespondingly.Therearealsomanysoftwareorappswhichusethisfunctiontoprotectouraccountsandprivateinformation.

Inthefollowingstage,ourtopicisaboutIntelligentCruiseControl.IntheComputerVision,withthecameraandimageprocessing,outtargetcancruiseautomaticallyandintelligently.Now,pleasewatchashortvideoaboutIntelligentCruiseControl.

Finally,wewilldosomeresearchaboutHuman-ComputerInterface.Peoplecaninteract[ɪntər'ækt]withcomputerwiththefunctionofComputerVision.Here,IwillgiveyouavideoaboutUsingKinectforWindowswithMATLAB.

Nextstage,IwillintroduceanewtheoryoranewsubjecttoyouwhichiscalledOpenCV.

Part3OpenCV

FirstIwilldoabriefintroductionaboutOpenCVandthenwewillpaymuchattentiontoImageProcessingandFaceDetectionbasedonOpenCV.

OpenCVisacross-platformcomputervisionlibrarywhichisbasedontheopensourceReleased.AndOpenCVcanrunonLinux,WindowsandMacOSoperatingsystems.Itislightweightbutwithhighefficient.OpenCVconsistsofaseriesofCfunctionsandafewC++classes.Whatismore,OpenCValsoprovidesaninterfaceofPython,Ruby['ruːbɪ],MATLABandotherlanguages​​.Furthermore,manycommonalgorithmsforimageprocessingandcomputervisionarealsorealizedontheplatformofOpenCV.

Inthefollowingstage,wewillpayattentiontoImageProcessing.

First,letustalksomethingaboutsmoothingofimage.Smoothing,alsocalledblurring[blɝ]模糊,isasimpleandfrequentlyusedimageprocessingoperation.Therearemanyreasonsforsmoothing,butitisusuallydonetoreducenoiseorcameraartifacts.Smoothingisalsoimportantwhenwewishtoreducetheresolution[rezə'luːʃ（ə）n]分辨率ofanimageinaprincipled['prɪnsɪp（ə）ld]有原则的way.OpenCVoffersfivedifferentsmoothingoperationsatthistime.Allofthemaresupportedthroughonefunction,cvSmooth（）,whichtakesourdesiredformofsmoothingasanargument.

voidcvSmooth（

constCvArr*src,

CvArr*dst,

intsmoothtype=CV_GAUSSIAN,

intparam1=3,

intparam2=0,

doubleparam3=0,

doubleparam4=0

）;

Thesrcanddstargumentsaretheusualsourceanddestinationforthesmoothoperation.Thispicturescanreflectthefunctionofsmoothing.Andhereisanexampleofsmoothing.

Second,letustalksomethingaboutresizeofimage.Weoftenencounteranimageofsomesizethatwewouldliketoconverttoanimageofsomeothersize.Wemaywanttoupsize（zoomin）ordownsize（zoomout）theimage;wecanaccomplisheithertaskbyusingcvResize（）.Thisfunctionwillfitthesourceimageexactlytothedestinationimagesize.

voidcvResize（

constCvArr*src,

CvArr*dst,

intinterpolation=CV_INTER_LINEAR

）;

IftheROIissetinthesourceimagethenthatROIwillberesizedtofitinthedestinationimage.Likewise,ifanROIissetinthedestinationimagethenthesourcewillberesizedtofitintotheROI.Andhereisanexampleofresize.

Third,letustalksomethingaboutthresholdofimage.Frequentlywehavedonemanylayersofprocessingstepsandwanteithertomakeafinaldecisionaboutthepixels['piksəl]像素inanimageortocategorically[,kæti'gɔrikli]绝对地rejectthosepixelsbeloworabovesomevaluewhilekeepingtheothers.TheOpenCVfunctioncvThreshold（）accomplishesthesetasks.

doublecvThreshold（

CvArr*src,

CvArr*dst,

doublethreshold,

doublemax_value,

intthreshold_type

）;

Thebasicideaisthatanarrayisgiven,alongwithathreshold,andthensomethinghappenstoeveryelementofthearraydependingonwhetheritisbeloworabovethethreshold.

Andhereisanexampleofthreshold.

Finally,IwilldosomeintroductionaboutFaceDetection.FaceDetectioncandivideintofourparts:

Facedetection,Facepreprocessing,CollectandlearnfacesandFacerecognition.Inthestageoffacedetection,thecomputershalldetectwhetherthetargetisafaceornot.Then,facepreprocessing,withthefunctionsofOpenCV，thecomputercanextractthefeatureoftheface.Inthenextstage---collectandlearnfaces,thecomputershallcollectandlearnlotsoffeaturesofface.Finally,withthepreviouspreparation,thecomputercandetectorrecognizemostoffacesinanewimage.HerearethesourcecodeofFaceDetectionbasedonOpenCVwhichusedthelanguageofJava.

MainFunction

Test

Forexample,ifIjuststandinfrontofacameraandwiththefunctionofOpenCV,thecomputercandetectthatthisisafaceandaperson’sface,andnotabook.AndthisisaresultofFaceDetection.Thecomputerdetectsthatthereareeightfacesinthepicture.However,infacttherearesevenfacesinthepicture.Itisobviouslythatthecomputeralsodetectsthebookasaperson’sfaceandmakeamistake.

Attheendofmypresentation,wewilltalkaboutthedevelopmentofComputerVisionforafewminutes.

Part4DevelopmentofComputerVision

Asascienceaboutstudyinghowtomakeamachineto‘see’,ComputerVisonisbecomingamoreandmoreimportantroleinourlife.Asmyopinion,3Dreconstruction,especiallytheapplicationofKinect,RecognitionandTrackingwilldevelopfastandbecomemoreandmoreimportantinthefuture.

Ok.Thatisall.Thankyouforyourattention.Anyquestion?