简介:Weldpoolcontainssignificantinformationabouttheweldingprocess.TheweldpoolimagesofMAGweldingaredetectedbyLaserStrobesystem.AnalgorithmforextractingweldpooledgeisproposedaccordingtothecharacteristicsofMAGweldpoolimages.Themaximumweldpoollengthandwidtharecalculated.ThemeasurementdatacanbeusedtoverifytheresultsofweldingprocesssimulationandtoprovideagoodfoundationforautomaticcontrolofMAGweldingprocess.
简介:Imagesensorhasbeenoneofthekeytechnologiesinintellectualizedroboticswelding.Edgedetectionplaysanimportantrolewhenthevisiontechnologyisappliedinintellectualizedweldingroboticstechnologies.Thereareallkindsofnoisesinweldingenvironment.Thealgorithmsincommonusecannotbeappliedtotherecognitionofweldingenvironmentdirectly.Theedgeofimagescanbefellintofourtypes.Theweldimagesareclassifiedbythecharacteristicofweldingenvironmentinthispaper.Thispaperanalyzessomealgorithmsofedgedetectionaccordingtothecharacterofweldingimage,somerelativeadvantagesanddisadvantagesarepointedoutwhenthesealgorithmsareusedinthisfield,andsomesuggestionsaregiven.Thefeatureextractionofweldseamandweldpoolaretwotypicalproblemsintherealizationofintellectualizedwelding.Theiredgefeaturesareextractedandtheresultsshowtheapplicabilityofdifferentedgedetectors.Thetradeoffbetweenprecisionandcalculatedtimeisalsoconsideredfordifferentapplication.
简介:ThelateralwaveinultrasonicTOFD(timeofflightdiffraction)imagehasatailintransittime,whichdisturbsthedetectionandevaluationofshallowwelddefect.Meanwhile,thelateralwaveandback-wallechothatactasbackgroundaddredundantdataindigitalimageprocessing.Inordertoseparatedefectwavefromlateralwaveandpreparethewayforfollowingimageprocessing,analgorithmofbackgroundremovalmethodnamedasmean-subtractionisdeveloped.Basedonthis,animprovedmethodbystatisticoftheenergydistributionintheimageisproposed.Theresultsshowthatbychoosingproperthresholdvalueaccordingtotheaxialenergydistributionoftheimage,thebackgroundcanberemovedautomaticallyandthedefectsectionbecomespredominant.Meanwhile,diffractivewaveofshallowwelddefectcanbeseparatedfromlateralwaveeffectively.
简介:Duetothedisturbancesofspatters,dustsandstrongarclight,itisdifficulttodetectthemoltenpooledgeandtheweldlinelocationinCO2weldingprocesses.Themedianfilteringandself-multiplicationwasemployedtopreprocesstheimageoftheCO2weldinginordertodetecteffectivelytheedgeofmoltenpoolandthelocationofweldline.TheB-splinewaveletalgorithmhasbeeninvestigated,theinfluenceofdifferentscalesandthresholdsontheresultsoftheedgedetectionhavebeencomparedandanalyzed.TheexperimentalresultsshowthatbetterperformancetoextracttheedgeofthemoltenpoolandthelocationofweldlinecanbeobtainedbyusingtheB-splinewavelettransform.Theproposededgedetectionapproachcanbefurtherappliedtothecontrolofmoltendepthandtheseamtracking.