Manualmonitoringandseamtrackingthroughwatchingweldpoolimagesinreal-time,bynakedeyesorbyindustrialTV,areexperience-depended,subjective,laborintensive,andsometimesbiased.Soitisnecessarytorealizetheautomationofcomputer-aidedseamtracking.APAW(plasmaarcwelding)seamtrackingsystemwasdeveloped,whichsensesthemoltenpoolandtheseaminoneframebyavisionsensor,andthendetectstheseamdeviationtoadjusttheworkpiecemotionadaptivelytotheseampositionsensedbyvisionsensor.Anovelmoltenpoolareaimage-processingalgorithmbasedonmachinevisionwasproposed.Thealgorithmprocesseseachimageatthespeedof20frames/secondinreal-timetoextractthreefeaturevariablestogettheseamdeviation.Itisprovedexperimentallythatthealgorithmisveryfastandeffective.Issuesrelatedtothealgorithmarealsodiscussed.
