|
چکیده
|
Identifying wearmechanismsfromwornsurfacesisacomplexandtediousprocessinvolvinghigh expertise.Wearscarsfromthecontactsurfacescanactasapotentialindicatorofwearprocess undergone bythematerialduringthecourseofsurfaceinteraction.Currently,wearmechanismanalysis is performedsubjectivelyandusuallyonlybytheauthor(s)ofthestudy.Theaimofthispaperisto develop imageprocessingtechniquestoquantifytwowearmechanisms:abrasionandmicro-pitting.To characterizewearmechanisms,therequiredsurfacemorphologywasproducedusingweartestingof polymer inatwin-disctestrig.Themicrographsofpolymercontactsurfaceswereacquiredusinga conventionalopticalmicroscope.Subjectivequalityscoreswerecollectedbyastructuredhuman observer studytoquantifytheseverityofmicropittingandabrasion.Thesescoreswereusedtovalidate objectivescoresobtainedusingimageprocessingtechniques.Severalimageprocessingtechniquesare proposed todetectandquantifymicro-pittingandabrasionmechanisms.Thesetechniquesconsistof localandglobalthresholdingsegmentationwithandwithoutunevenilluminationcompensation, granulometrybybinaryopeningandgranulometrybygray-scaleclosing.Theproposedimageprocessing analysisrevealstheseverityofabrasionandpittingmechanisms(whentheyaredominant)whichagree wellwiththemeanopinionscores(MOS)givenbyobservers.Forabrasion,thecorrelationcoefficients betweenobjectivescoresobtainedbydifferentimageprocessingtechniquesandsubjectivevaluesshowa highlinearcorrelation.Forpitting,weclaimthatobjectivevaluesobtainedfromthenewmethods, distinguishthepittingmechanism.Thisstudyisaproofofconcepttowardsautomatedidentificationand quantificationofdifferentwearmechanismsasareplacementforhumanobservers.
|