Visione computerizzata

Definition

Computervisionisasimulationofbiologicalvisionusingcomputersandrelatedequipment.Itsmaintaskistoobtainthree-dimensionalinformationofthecorrespondingscenebyprocessingthecollectedpicturesorvideos,justlikehumansandmanyotherkindsofcreaturesdoeveryday.

Computervisionisasubjectofhowtousecamerasandcomputerstoobtainthedataandinformationweneedaboutthesubject.Toputitvividly,itistoinstalleyes(cameras)andbrains(algorithms)onthecomputersothatthecomputercanperceivetheenvironment.TheChineseidiom"seeingisbelieving"andthewesternsaying"Onepictureisworthtenthousandwords"expresstheimportanceofvisiontomankind.Itisnotdifficulttoimaginehowbroadtheapplicationprospectsofmachineswithvisioncanbe.

Computervisionisnotonlyanengineeringfield,butalsoachallengingandimportantresearchfieldinthescientificfield.Computervisionisacomprehensivediscipline,ithasattractedresearchersfromvariousdisciplinestoparticipateinitsresearch.Theseincludecomputerscienceandengineering,signalprocessing,physics,appliedmathematicsandstatistics,neurophysiologyandcognitivescience.

Analysis

Visionisanintegralpartofvariousintelligent/autonomoussystemsinvariousapplicationfields,suchasmanufacturing,inspection,documentanalysis,medicaldiagnosis,andmilitary.Becauseofitsimportance,someadvancedcountries,suchastheUnitedStates,listcomputervisionresearchasamajorbasicprobleminscienceandengineeringthathasawide-rangingimpactoneconomyandscience,theso-calledgrandchallenge.Thechallengeofcomputervisionistodevelopvisioncapabilitiescomparabletohumansforcomputersandrobots.Machinevisionrequiresimagesignals,textureandcolormodeling,geometricprocessingandreasoning,andobjectmodeling.Acapablevisionsystemshouldtightlyintegratealltheseprocesses.Asadiscipline,computervisionbeganintheearly1960s,butmanyimportantadvancesinthebasicresearchofcomputervisionweremadeinthe1980s.Computervisioniscloselyrelatedtohumanvision.Acorrectunderstandingofhumanvisionwillbeverybeneficialtotheresearchofcomputervision.Forthiswewillfirstintroducehumanvision.

Principium

Computervisionistheuseofvariousimagingsystemsinsteadofvisualorgansasinput-sensitivemeans,andcomputersinsteadofthebraintocompleteprocessingandinterpretation.Theultimateresearchgoalofcomputervisionistoenablecomputerstoobserveandunderstandtheworldthroughvisionlikehumans,andhavetheabilitytoadapttotheenvironmentautonomously.Agoalthatcanonlybeachievedafterlong-termefforts.Therefore,beforeachievingthefinalgoal,themid-termgoalofpeople'seffortsistoestablishavisionsystemthatcancompletecertaintasksbasedonacertaindegreeofintelligencewithvisualsensitivityandfeedback.Forexample,animportantapplicationareaof​​computervisionisthevisualnavigationofautonomousvehicles.Thereisnoconditiontorealizeasystemthatcanrecognizeandunderstandanyenvironmentandcompleteautonomousnavigationlikehumans.Therefore,theresearchgoalofpeople'seffortsistoachieveavisualassisteddrivingsystemthathasroadtrackingcapabilitiesonexpresswaysandcanavoidcollisionswithvehiclesinfront.Thepointtobepointedouthereisthatinthecomputervisionsystem,thecomputerreplacesthehumanbrain,butitdoesnotmeanthatthecomputermustcompletetheprocessingofvisualinformationaccordingtothemethodofhumanvision.Computervisioncanandshouldprocessvisualinformationaccordingtothecharacteristicsofthecomputersystem.However,thehumanvisualsystemisbyfarthemostpowerfulandcompletevisualsystemknowntopeople.Asyouwillseeinthefollowingchapters,thestudyofhumanvisualprocessingmechanismswillprovideinspirationandguidanceforcomputervisionresearch.Therefore,thecomputerinformationprocessingmethodisusedtostudythemechanismofhumanvisionandestablishthecalculationtheoryofhumanvision.ResearchinthisareaiscalledComputationalVision.Computationalvisioncanbeconsideredasaresearchfieldincomputervision.

Related

Therearemanydisciplineswhoseresearchgoalsaresimilartoorrelatedtocomputervision.Thesesubjectsincludeimageprocessing,patternrecognitionorimagerecognition,sceneanalysis,imageunderstanding,etc.Computervisionincludesimageprocessingandpatternrecognition.Inaddition,italsoincludesthedescriptionofspatialshapes,geometricmodeling,andtheprocessofrecognition.Realizingimageunderstandingistheultimategoalofcomputervision.

Imageprocessing

Imageprocessingtechnologyconvertstheinputimageintoanotherimagewithdesiredcharacteristics.Forexample,theoutputimagecanbeprocessedtohaveahighersignal-to-noiseratio,orenhancedprocessingcanbeusedtohighlightthedetailsoftheimagetofacilitateinspectionbytheoperator.Incomputervisionresearch,imageprocessingtechnologyisoftenusedforpreprocessingandfeatureextraction.

Patternrecognition

Patternrecognitiontechnologydividesimagesintopredeterminedcategoriesbasedonthestatisticalcharacteristicsorstructuralinformationextractedfromtheimage.Forexample,textrecognitionorfingerprintrecognition.Incomputervision,patternrecognitiontechnologyisoftenusedtoidentifyandclassifycertainpartsofanimage,suchassegmentedregions.

Imageunderstanding

Givenanimage,theimageunderstandingprogramnotonlydescribestheimageitself,butalsodescribesandinterpretsthescenerepresentedbytheimage,inordertomakeananalysisofthecontentrepresentedbytheimage.Decide.Intheearlydaysofartificialintelligencevisionresearch,thetermsceneanalysiswasoftenusedtoemphasizethedifferencebetweentwo-dimensionalimagesandthree-dimensionalscenes.Inadditiontocompleximageprocessing,imageunderstandingalsorequiresknowledgeaboutthephysicallawsofsceneimagingandknowledgerelatedtothecontentofthescene.

Whenestablishingacomputervisionsystem,itisnecessarytousetherelevanttechnologiesintheabovedisciplines,butthecontentofcomputervisionresearchismoreextensivethanthesedisciplines.Theresearchofcomputervisioniscloselyrelatedtotheresearchofhumanvision.Inordertoachievethegoalofestablishingageneral-purposecomputervisionsystemsimilartothehumanvisionsystem,itisnecessarytoestablishacomputertheoryofhumanvision.

Nunc et situ

Theoutstandingfeatureofthecomputervisionfieldisitsdiversityandimperfection.Pioneersinthisfieldcanbetracedbacktoearliertimes,butitwasnotuntilthelate1970swhentheperformanceofcomputerswasimprovedtohandlelarge-scaledatasuchasimagesthatcomputervisionreceivedformalattentionanddevelopment.However,thesedevelopmentsoftenoriginatefromtheneedsofotherdifferentfields,sowhatismeantby"computervisionproblems"hasneverbeenformallydefined.Naturally,thereisnoformulaforhow"computervisionproblems"shouldbesolved.

Nevertheless,peoplehavebeguntomastersomeofthemethodstosolvespecificcomputervisiontasks.Unfortunately,thesemethodsareusuallyonlyapplicabletoagroupofnarrowtargets(suchas:faces,fingerprints,text,etc.),sotheycannotbeWidelyusedindifferentoccasions.

Theapplicationofthesemethodsisusuallyacomponentofsomelarge-scalesystemsthatsolvecomplexproblems(suchasmedicalimageprocessing,qualitycontrolandmeasurementinindustrialmanufacturing).Inmostpracticalapplicationsofcomputervision,computersarepresettosolvespecifictasks.However,methodsbasedonmachinelearningarebecomingmoreandmorepopular.Oncetheresearchofmachinelearningisfurtherdeveloped,thefuture"generalpurpose"computervisionapplicationsmaybeabletocometrue.

Oneofthemainissuesstudiedbyartificialintelligenceis:howtomakethesystemhave"planning"and"decision-makingcapabilities"?Soastomakeitcompleteaspecifictechnicalaction(forexample:movearobotthroughaspecificenvironment).Thisproblemiscloselyrelatedtothecomputervisionproblem.Here,thecomputervisionsystemactsasaperceptron,providinginformationfordecision-making.Otherresearchdirectionsincludepatternrecognitionandmachinelearning(whichalsobelongtothefieldofartificialintelligence,buthaveanimportantconnectionwithcomputervision).Asaresult,computervisionisoftenregardedasabranchofartificialintelligenceandcomputerscience.

Physicsisanotherfieldthathasanimportantconnectionwithcomputervision.

Thegoalofcomputervisionistofullyunderstandtheelectromagneticwaves-mainlyvisiblelightandinfraredlight-theimageformedbythereflectionofthesurfaceoftheobject,andthisprocessisbasedonopticalphysicsandsolid-statephysics.Somecutting-edgeimageperceptionsystemswillevenbeappliedtoquantummechanicstheorytoanalyzetherealworldrepresentedbyimages.Atthesametime,manymeasurementproblemsinphysicscanalsobesolvedbycomputervision,suchasfluidmotion.Becauseofthis,computervisioncanalsobeseenasanextensionofphysics.

Anotherimportantfieldisneurobiology,especiallythepartofthebiologicalvisualsystem.

Throughoutthe20thcentury,humanshaveconductedextensivestudiesontheeyes,neurons,andbraintissuesofvariousanimalsrelatedtovisualstimulation.Thesestudieshaveledtosome"natural"Thedescriptionofhowthevisualsystemworks(althoughitisstillabitrough)hasalsoformedasub-fieldofcomputervision-peopletrytobuildartificialsystemsthatcansimulatethevisualoperationsoflivingbeingswithvaryingdegreesofcomplexity.Atthesametime,inthefieldofcomputervision,somemethodsbasedonmachinelearningalsorefertosomebiologicalmechanisms.

Anotherrelatedfieldofcomputervisionissignalprocessing.Manyprocessingmethodsrelatedtounitvariablesignals,especiallytheprocessingoftime-varyingsignals,cannaturallybeextendedtotheprocessingmethodsofbinaryvariablesignalsormultivariatesignalsincomputervision.However,duetotheuniquepropertiesofimagedata,manymethodsdevelopedincomputervisioncannotfindacorrespondingversionintheunitsignalprocessingmethod.Oneofthemaincharacteristicsofthesemethodsistheirnon-linearityandthemulti-dimensionalityofimageinformation.Theabovetwopoints,aspartofcomputervision,formaspecialresearchdirectioninsignalprocessing.

Inadditiontothefieldsmentionedabove,manyresearchtopicscanalsobetreatedaspurelymathematicalproblems.Forexample,manyproblemsincomputervisionarebasedonstatistics,optimizationtheory,andgeometry.

Howtoimplementexistingmethodsthroughvarioussoftwareandhardware,orhowtomodifythesemethods,soastoobtainreasonableexecutionspeedwithoutlosingsufficientaccuracy,isthemainissueinthefieldofcomputervisiontoday.Subject.

Applicationem

Mankindisenteringtheinformationage,andcomputerswillincreasinglyenteralmostallfields.Ontheonehand,morepeoplewithoutprofessionalcomputertrainingalsoneedtousecomputers.Ontheotherhand,thefunctionsofcomputersaregettingstrongerandstronger,andthemethodsofusingthemaregettingmoreandmorecomplicated.Thiscreatesasharpcontradictionbetweentheflexibilityofpeopleinconversationandcommunicationandthestrictnessandrigidityrequiredwhenusingcomputers.Humanscanexchangeinformationwiththeoutsideworldthroughvision,hearing,andlanguage,andcanexpressthesamemeaningindifferentways.However,computersarerequiredtowriteprogramsstrictlyinaccordancewithvariousprogramminglanguages,sothatcomputerscanrun.Inordertoenablemorepeopletousecomplexcomputers,itisnecessarytochangethepastsituationwherepeopleadapttocomputersandmemorizecomputerusagerulesbyrote.Instead,letthecomputeradapttopeople'shabitsandrequirements,andexchangeinformationwithpeopleinthewaypeopleareusedto,thatis,letthecomputerhavetheabilitytosee,hear,andspeak.Atthistime,thecomputermusthavetheabilityoflogicalreasoninganddecision-making.Acomputerwiththeabovecapabilitiesisanintelligentcomputer.

Intelligentcomputersnotonlymakecomputersmoreconvenientforpeopletouse,butatthesametime,ifsuchcomputersareusedtocontrolvariousautomationdevices,especiallyintelligentrobots,theseautomationsystemsandintelligentrobotscanadapttotheenvironment,andTheabilitytomakedecisionsindependently.Thiscanreplacepeople'sheavyworkonvariousoccasions,orreplacepeopletocompletetasksinvariousdangerousandharshenvironments.

Applicationemsrangefromtasks,suchasindustrialmachinevisionsystems,forexample,inspectionofbottlesontheproductionlinetoacceleratethrough,researchintoartificialintelligenceandcomputersorrobots,whichcanunderstandtheworldaroundthem.Thereisasignificantoverlapinthefieldsofcomputervisionandmachinevision.Computervisioninvolvesthecoretechnologyusedinautomatedimageanalysisinmanyfields.Machinevisionusuallyreferstoaprocessthatcombinesautomaticimageanalysiswithothermethodsandtechnologiestoprovideautomaticdetectionandrobotguidanceinindustrialapplications.Inmanycomputervisionapplications,computersarepre-programmedtosolvespecifictasks,butlearning-basedmethodsarenowbecomingmoreandmorecommon.Examplesofcomputervisionapplicationsincludethoseusedinsystems:

(1) Continere processus, suchasanindustrialrobot;

(2) Navigatio, sicut carrus autonomus aut mobilerobots;

(3) Events detecta, such asveillance and people counting;

(4) Organisinginformation, forexample, indexdata basibus for images and images sequences;

(5)Modelingobjectsorenvironments,suchasmedicalimageanalysissystemsorterrainmodels;

(6) Commercium, exempli causa, cum inputandae vice-computationis humanitatis negotium est;

(7) Automaticdetection, for example, inmanufaciens applicationes.

Themostprominentapplicationareasaremedicalcomputervisionandmedicalimageprocessing.Thefeatureinformationofthisareaisextractedfromtheimagedataforthepurposeofmedicaldiagnosisofthepatient.Usually,theimagedataisintheformofmicroscopeimages,X-rayimages,angiographyimages,ultrasoundimagesandtomographicimages.Anexampleoftheinformationthatcanbeextractedfromsuchimagedataisthedetectionoftumors,atherosclerosisorothermalignantchanges.Itcanalsobethesizeoftheorgan,bloodflow,etc.Thisfieldofapplicationalsosupportsthemeasurementofmedicalresearchbyprovidingnewinformation,forexample,onthestructureofthebrain,oraboutthequalityofmedicaltreatment.Theapplicationofcomputervisioninthemedicalfieldalsoincludesenhancingimagesthatareinterpretedbyhumans,suchasultrasoundimagesorX-rayimages,toreducetheeffectsofnoise.

Thesecondapplicationareaof​​computervisionisinindustry,sometimescalledmachinevision,whereinformationisextractedtosupportthepurposeofthemanufacturingprocess.Anexampleisqualitycontrol,wheretheinformationorfinalproductisautomaticallydetectedinordertofinddefects.Anotherexampleisthatthepositionanddetailorientationbeingpickeduparemeasuredbytheroboticarm.Machinevisionisalsousedextensivelyintheprocessofagriculture,frombulkmaterials,thisprocessiscalledtheremovalofunwantedthings,opticalsortingoffood.

Militaryapplicationsareprobablyoneofthelargestareasofcomputervision.Themostobviousexampleisthedetectionofenemysoldiersorvehiclesandmissileguidance.Moreadvancedsystemsguidethemissiletotheareawherethemissileissent,ratherthanaspecifictarget,andmakeaselectionwhenthemissilereachesthetargetintheareabasedonlocallyacquiredimagedata.Modernmilitaryconcepts,suchas"battlefieldperception",meanthatvarioussensors,includingimagesensors,provideawealthofrelevantcombatscenariosthatcanbeusedtosupportstrategicdecision-makinginformation.Inthiscase,automaticdataprocessingisusedtoreducecomplexityandfuseinformationfrommultiplesensorstoimprovereliability.

Anewerapplicationareaisautonomousvehicles,whichincludediving,landvehicles(smallrobotswithwheels,carsortrucks),aerialworkvehiclesandunmannedaerialvehicles(UAV).Thelevelofautonomyrangesfromcompletelyindependent(unmanned)vehiclestocars,wherecomputervision-basedsystemssupportdriverprogramsorexperimentsindifferentsituations.Afullyautonomouscarusuallyusescomputervisiontonavigatewhenitknowswhereitis,ortheenvironmentusedforproduction(mapSLAM)andfordetectingobstacles.Itcanalsobeusedtodetectspecificeventsforspecifictasks,forexample,aUAVlookingforforestfires.Examplesofsupportsystemsarecarsinobstaclewarningsystems,andautonomouslandingsystemsforaircraft.Severalautomakershavedemonstratedthesystem'sautonomousdrivingofcars,butthetechnologyhasnotreachedacertainlevelbeforeitcanbeputonthemarket.Thereareplentyofexamplesofmilitaryautonomousmodels,fromadvancedmissiles,unmannedaerialvehiclesforreconnaissancemissionsormissileguidance.Spaceexplorationisalreadyusingcomputervision,autonomousvehiclessuchasNASA’sMarsExplorationRoverandtheEuropeanSpaceAgency’sExoMarsMarsRover.

Otherapplicationareasinclude:

(1) Movies and broadcasts that support the production of visualeffects, forexemple,cameratracking(motionmatching).

(2) Cras.

Similitudines anddifferences

Computervision,imageprocessing,imageanalysis,robotvisionandmachinevisionarecloselyrelateddisciplines.Ifyouopenthetextbookswiththeabovenames,youwillfindthattheyhaveaconsiderableoverlapintechnologyandapplicationareas.Thisshowsthatthebasictheoriesofthesedisciplinesareroughlythesame,anditevenmakespeoplesuspectthattheyarethesamedisciplineswithdifferentnames.

Computer vision

However,variousresearchinstitutions,academicjournals,conferences,andcompaniesoftenclassifythemselvesasaparticularfield,soavarietyofcharacteristicsthatdistinguishthesedisciplineshavebeenbroughtup.Amethodofdistinctionwillbegivenbelow,althoughitcannotbesaidthatthismethodofdistinctioniscompletelyaccurate.

Theresearchobjectofcomputervisionismainlyathree-dimensionalscenemappedtoasingleormultipleimages,suchasthereconstructionofathree-dimensionalscene.Theresearchofcomputervisionislargelyfocusedonthecontentoftheimage.

Theresearchobjectsofimageprocessingandimageanalysisaremainlytwo-dimensionalimages,whichrealizeimagetransformation,especiallyforpixel-leveloperations,suchasimagecontrastimprovement,edgeextraction,denoisingandgeometrictransformationssuchasimagerotation.Thisfeatureshowsthattheresearchcontentofimageprocessingorimageanalysishasnothingtodowiththespecificcontentoftheimage.

Machinevisionmainlyreferstothevisualresearchintheindustrialfield,suchasthevisionofautonomousrobots,andthevisionforinspectionandmeasurement.Thisshowsthatinthisfield,throughsoftwareandhardware,imageperceptionandcontroltheoryisoftencloselycombinedwithimageprocessingtoachieveefficientrobotcontrolorvariousreal-timeoperations.

Patternrecognitionusesvariousmethodstoextractinformationfromsignals,mainlyusingstatisticaltheories.Oneofthemaindirectionsinthisfieldistoextractinformationfromimagedata.

Thereisanotherfieldcalledimagingtechnology.Theinitialresearchcontentinthisfieldismainlytomakeimages,butsometimesalsoinvolvesimageanalysisandprocessing.Forexample,medicalimagingincludesalargenumberofimageanalysisinthemedicalfield.

Forallthesefields,apossibleprocessisthatyouworkinacomputervisionlaboratory,youareengagedinimageprocessingatwork,andfinallysolvetheproblemsinthefieldofmachinevision,andthenpublishyourresultsinAtthemeetingofpatternrecognition.

Problemata

Almosteveryspecificapplicationofcomputervisiontechnologymustsolveaseriesofthesameproblems.Theseclassicproblemsinclude:

Recognitio

Acomputervision,imageprocessingandmachinevisioncommonclassicproblemistodeterminewhetherasetofimagedatacontainsaspecificObject,imagefeatureormovementstate.Thisproblemcanusuallybesolvedautomaticallybyamachine,butsofar,thereisnosinglemethodthatcandetermineawiderangeofsituations:recognizeanyobjectinanyenvironment.Theexistingtechnologycanandcanonlywellsolvetherecognitionofspecifictargets,suchassimplegeometricpatternrecognition,facerecognition,printedorhandwrittendocumentrecognition,orvehiclerecognition.Andtheserecognitionsneedtohavespecifiedlighting,backgroundandtargetposturerequirementsinaspecificenvironment.

Generalrecognitionhasevolvedintoseveralslightlydifferentconceptsondifferentoccasions:

Recognitio(narrowsense):Foroneormorepre-definedorlearnedObjectsorobjectsarerecognized,andtheirtwo-dimensionalpositionorthree-dimensionalpostureisusuallyprovidedduringtherecognitionprocess.

Identification:Identifythesingleobjectitself.Forexample:therecognitionofacertainface,therecognitionofacertainfingerprint.

Monitoring:Discoverspecificsituationcontentfromimages.Forexample:thediscoveryofabnormalskillsincellsortissuesinmedicine,andthediscoveryofpassingvehiclesbytrafficmonitoringequipment.Monitoringisoftentodiscoverspecialareasintheimagethroughsimpleimageprocessing,whichprovidesastartingpointforsubsequentmorecomplexoperations.

Pluresspecificapplicationdirectionsidentified:

Content-basedimageextraction:Findallpicturescontainingspecifiedcontentinahugeimagecollection.Thespecifiedcontentcantakemanyforms,suchasaredroughlycircularpattern,orabicycle.Thesearchforthelatterkindofcontenthereisobviouslymorecomplicatedthantheformer,becausetheformerdescribesalow-levelintuitivevisualfeature,whilethelatterinvolvesanabstractconcept(orhigh-levelvisualfeature).Thatis,"bicycle",theobviouspointisthattheappearanceofthebicycleisnotfixed.

Poseevaluation:Evaluationofthepositionordirectionofanobjectrelativetothecamera.Forexample:theassessmentofthepostureandpositionoftheroboticarm.

Opticalcharacterrecognitionrecognizesanddiscriminatesprintedorhandwrittentextinanimage,andtheusualoutputistoconvertitintoaneasy-to-editdocumentform.

Motus

Themonitoringofobjectmotionbasedonsequenceimagesincludesmanytypes,suchas:

Selfmotion:monitorthethree-dimensionalrigidmotionofthecamera.

Imagetracking:Trackmovingobjects.

SceneReconstruction

Giventwoormoreimagesoravideoofascene,scenereconstructionseekstobuildacomputermodel/three-dimensionalmodelofthescene.Thesimplestcaseistogenerateasetofpointsinthree-dimensionalspace.Inmorecomplexsituations,acompletethree-dimensionalsurfacemodelwillbebuilt.

Imagerestoration

Thegoalofimagerestorationistoremovenoiseintheimage,suchasinstrumentnoise,blur,etc.

Systema

Thestructureofthecomputervisionsystemlargelydependsonitsspecificapplicationdirection.Someworkindependentlyandareusedtosolvespecificmeasurementorinspectionproblems;someappearasapartofalargecomplexsystem,suchasworkingwithmechanicalcontrolsystems,databasesystems,andman-machineinterfacedevices.Thespecificimplementationmethodofthecomputervisionsystemisalsodeterminedbyitsfunction-whetheritisfixedinadvanceorisautomaticallylearnedandadjustedduringoperation.However,therearesomefunctionsthatalmosteverycomputersystemneeds:

Imageacquisition

Adigitalimageisproducedbyoneormoreimagesensors,hereThesensorcanbeavarietyofphotosensitivecameras,includingremotesensingequipment,X-raytomography,radar,ultrasonicreceivers,andsoon.Dependingonthedifferentperceptrons,thegeneratedpicturecanbeanordinarytwo-dimensionalimage,athree-dimensionalimagegrouporanimagesequence.Thepixelvalueofthepictureoftencorrespondstotheintensityoflightinoneormorespectralbands(grayscaleorcolorimage),butitcanalsoberelatedtovariousphysicaldata,suchasthedepthandabsorbanceofsoundwaves,electromagneticwavesornuclearmagneticresonanceOrreflectivity.

Preprocessing

Beforeimplementingspecificcomputervisionmethodsontheimagetoextractcertainspecificinformation,oneorsomepreprocessingisoftenusedtomaketheimagemeettherequirementsofsubsequentmethodsRequire.Forexample:

Sub-sampling toensure the correct imagecoordinates;

Smoothdenoisingtofilteroutthedevicenoiseintroducedbythesensor;

ImprovethecontrasttoensuretherealizationRelevantinformationcanbedetected;

Adjustthescalespacetomaketheimagestructuresuitableforlocalapplications.

Featureextraction

Extractfeaturesofvariouscomplexityfromtheimage.Forexample:

Linea, edgeextraction;

Localizedfeaturepointdetectionsuchascornerdetection,spotdetection;

MorecomplexfeaturesmayberelatedtotheimageThetextureshapeormovementisrelated.

Detectionsegmentation

Intheprocessofimageprocessing,itissometimesnecessarytosegmenttheimagetoextractvaluablepartsforsubsequentprocessing,suchas

screeningFeaturepoints;

Segmentthepartofoneormorepicturesthatcontainsaspecifictarget.

Advancedprocessing

Atthispoint,thedataoftenhasasmallamount,suchasthepartoftheimagethatisconsideredtocontainthetargetobjectafterpreviousprocessing.Theprocessingatthistimeincludes:

Verifywhetherthedataobtainedmeetstheprerequisiterequirements;

Estimatespecificcoefficients,suchasthetarget’sattitudeandvolume;

sort.

Advancedprocessinghasthemeaningofunderstandingimagecontent.Itisahigh-levelprocessingincomputervision.Itismainlybasedonimagesegmentationtounderstandthesegmentedimageblocks,suchasrecognitionandotheroperations..

Requisita

Theinfluenceoflightsourcelayoutneedstobecarefullyconsidered.

Select the correctlensgroup,considering themagnification, space, size, distortion...

Choosetherightcamera(CCD), considerans thefunctionem, speciem, stabilitatem, firmitatem ...

Visualsoftwaredevelopmentneedstorelyontheaccumulationofexperience,trymoreandthinkaboutthewaytosolvetheproblem.

Theultimategoalistocontinuouslyimprovetheaccuracyofcreationandshortentheprocessingtime.

finem.

Colloquium

Top

ICCV:InternationalColloquiumonComputerVision,InternationalComputerVisionColloquium

CVPR:InternationalColloquiumonComputerVisionandPatternRecognitio,InternationalColloquiumonComputerVisionandPatternRecognitio

ECCV:EuropeanColloquiumonComputerVision,EuropeanColloquiumonComputerVision

Melior

ICIP:InternationalColloquiumonImageProcessing,InternationalColloquiumonImageProcessing

BMVC:BritishMachineVisionColloquium,BritishMachineVisionColloquium

ICPR:InternationalColloquiumonPatternRecognitio,InternationalColloquiumonPatternRecognitio

ACCV:AsianColloquiumonComputerVision,AsianColloquiumonComputerVision

Journal

Top

PAMI:IEEETransactionsonPatternAnalysisandMachineIntelligence,IEEEPatternAnalysisJournalofMachineIntelligence

IJCV:InternationalJournalonComputerVision,InternationalJournalofComputerVision

Melior

TIP:IEEETransactionsonImageProcessing,IEEEImageProcessingMagazine

CVIU:ComputerVisionandImageUnderstanding,ComputerVisionandImageUnderstanding

PR:PatternRecognitio,PatternRecognitio

PRL:PatternRecognitioLetters,PatternRecognitioExpress

Related Articles
TOP