I'm not comfortable commissioning our UR5e bezel inspection from these labels. A fixed camera sees bright streaks move with small presentation changes on the reflective aluminium, yet the examples call them scratches. What should establish the physical references and capture conditions before we tune thresholds?
moving bezel streaks are already being used as scratch targets
PavelChan1049 · 2026年7月31日 23:49 UTC
主帖中文翻译暂不可用，以下显示英文原文；回复保留原文。 查看英文原文
13 条回复
Physical acceptance first. Can the pictured bezels still be individually identified and examined?
15分What changed between those pictures: seating, orientation, lighting, or several things? The streak moving is useful evidence, but it doesn't decide whether the surface also has a scratch.
6分Preserve the full images and their setup notes. Cropping to the bright region may hide a seating difference. Keep uncertain labels out of the accepted tuning set without deleting the examples.
24分I've seen passed test results lose all their limitations in a handover, so I'd put the reference status beside each image where the next person will actually find it
2分The bezels are retained. The labels were assigned while looking at the pictures, not from an agreed physical acceptance review. We have small presentation differences but haven't isolated which capture change causes which streak.
10分That leaves two questions open: the bezel's condition and the imaging setup's repeatability.
13分Sara, would you stop all imaging comparisons until quality finishes? I can see value in studying the seating now, provided nobody calls it a defect test.
17分No. Study repeatability now; defer defect-performance claims until the physical references are settled.
19分Use identified pieces and record one deliberate permitted setup change at a time for that study. The comparison should distinguish a setup effect from swapping to a different bezel.
14分Quality has now examined the retained pieces under an agreed view. We have an accepted surface and a confirmed scratch to carry into the capture comparison. The original picture labels weren't reliable enough to keep unchanged.
6分Thanks for coming back, Pavel; will the old labelled copies in the commissioning notes be corrected too?
20分Yes, I've marked the old examples superseded and linked the physical judgements. Capture comparison is still underway. I don't have a threshold recommendation or evidence over the wider finish range yet.
17分My earlier plate comparison reduced a highlight without losing the confirmed scratch. Useful locally, not a bezel result.
8分