You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Fixed: a white label on a dark box with the label edge one or two modules from the symbol was not read (the symbol merged with the dark background). At a 2-module edge the read rate went from 5 of 12 to 12 of 12 on synthetic frames, at 1 module from 0 to 8 of 12.
Fixed: a symbol touching the edge of the frame was never read (its finder pattern was smeared outside the frame); it is read now.
Fixed: a false read — a 40×40 symbol with text right next to two of its sides could decode to garbage when all error-correction capacity was spent on erasures. At least four syndromes are now always left to verify the result; multi-block symbols without pad codewords get the same margin as one-block ones. No false reads on 4000 frames without a valid code (text, QR, Code 128, fake and half symbols, overdamaged and mirrored symbols).
Dot-peen (DPM) marking reads markedly better: 8–12 of 12 synthetic frames instead of 0–8.
Checked frame by frame against 1.0.3 on our production-line sets: camera conveyor 8922 of 8959 frames (+2), pharmacy line 4011 of 4019 (unchanged); no frame lost, no string changed; per-frame time unchanged.