The bright bands move when we present the faceplate differently, yet the planning folder calls them scratches. Before commissioning, how do I give quality and imaging a reference set they can actually judge, including any real mark hidden by the glare?
Identify the retained faceplates and claimed regions, then obtain physical judgements under the appearance requirement. Keep capture and seating conditions linked to each view. Uncertain captions can remain diagnostic history, but they cannot be the answers the proposed inspection is trained to reproduce
Two retained pieces have definite image links. Quality is judging those. One old crop has no recoverable piece identity, so it's outside the confirmed set even though somebody circled a very convincing bright line.
Good separation. Will the new comparison retain full views and repeat presentations of each identified piece? A crop can locate a suspicious image feature without showing how the part was seated or whether the next view belongs to the same surface
Yes. Quality accepted one linked faceplate and found a reject groove on the other. We have full views locating both regions; the groove disappears under glare in one repeat. That failure is staying in the next imaging comparison, with no threshold chosen yet.
Then the physical references are firmer and the visibility failure is explicit. Keep the accepted bright finish in the same comparison, so revealing the groove does not quietly become rejecting every reflective faceplate