Multi-SKU Changeover With One Camera

One camera serving several products is a resolution compromise plus a changeover procedure. The field of view has to contain the largest product while the object-side pixel has to be small enough for the smallest measured feature on any product — and because the object-side pixel is the sensor pixel divided by one magnification, the smallest feature is measured with the same object-side pixel as everything else. That arithmetic, not the optics catalogue, decides whether a single station is feasible.

Sizing the compromise

Work in one direction: largest field required → magnification → object-side pixel → pixels across the smallest feature. With a 3.45 µm pixel and 0.14× the object-side pixel is about 24.6 µm, so a 0.1 mm feature spans only about four pixels — adequate for detection, marginal for measurement. If the tolerance needs ten pixels across that feature, the single-camera option fails on paper and the decision moves up a level: a second station, a zoom lens that changes magnification between product groups, or a higher-resolution sensor. The same arithmetic decides whether the height spread fits the depth of field, with focus at mid-height and a working rule of spread ≤ 0.7 × DOF.

What a changeover really costs

A hardware lens swap costs line downtime, refitting and re-establishing the working distance, refocusing, recalibration, a first-article inspection and the documentation — and it repeats every time. A recipe change on a motorized lens costs seconds and no mechanical intervention, but still needs first-article verification. When changeovers are frequent, the payback of a motorized zoom or a second station is driven by those hours rather than by optical quality; write the comparison with the changeover frequency you actually have, and include the hidden items, or the comparison flatters the manual approach.

Recipes, fixtures and the risk of mixing parts

Treat recipes as controlled documents: zoom, focus, aperture, scale factor, region of interest and illumination settings belong to a revision, and any change should trigger re-verification against a reference feature. Fixtures should guarantee repeatable presentation — indexed adjustable stops or quick-change inserts seating against a hard datum — because a presentation error appears directly in the measurement and looks like process variation. Finally, identify the part automatically (barcode, RFID or vision classification) and refuse to run without a match, so that product A is never measured with the recipe for product B.

Related reading: parts with different heights, industrial lens selection, motorized zoom lens selection, and the machine vision lens range. The questions on resolution compromise, fixtures and recipe control are answered below — send the SKU list and the tolerances and we will check whether one station covers them.