Why Vision System Installations Should Be Done With Two People

We have a rule in the shop that costs us money on every service quote and saves us money on every service call: vision installations get two people. One at the HMI, one at the camera. Customers sometimes push back on the second set of travel expenses, and we understand why - bolting a camera to a bracket does not look like a two-person job. It isn't. The two-person job is everything that happens after the bracket.

The geometry problem

Focusing and aiming an industrial camera is a live, two-variable optimization, and the two variables live in different places. The lens is at the machine. The image is on a monitor, which on a real production line is somewhere else - across the aisle, up on the mezzanine, or inside a panel two guard doors away. The feedback loop for every adjustment runs through a human walking between them.

Watch what actually happens when one person commissions a camera alone. They tweak the focus ring, walk to the monitor, look, walk back, tweak again. Each round trip costs a minute or two. Somewhere around the third or fourth trip, a very human thing occurs: the image gets called good. Not because it is good, but because the cost of one more iteration finally exceeds the installer's belief that the next tweak will matter. Focus, aperture, working distance, polarizer angle, and aim all get frozen at "good enough to stop walking."

With a second person at the monitor, the iteration cost drops to the length of a sentence. "Quarter turn clockwise. Too far. Back an eighth." Twenty iterations happen in the time the solo installer spends on three. The final image is not slightly better - it is a different class of result, because the optimization actually ran to completion instead of being abandoned.

The honesty problem

The subtler reason is the one the teaser for this article promised, and it is about judgment, not legs.

The person who just spent ten minutes adjusting a lens is the wrong person to grade the image, for the same reason a writer is the wrong person to proofread their own paragraph. They see the image they intend, not the image on the screen. They know how much effort the current setting cost, and that effort quietly argues for acceptance. Every commissioning engineer has felt the pull of "it will probably be fine" at hour eleven of a startup - and the person at the camera feels it hardest, because they are the one who has to climb back up the ladder if it isn't.

The person at the HMI has no such investment. They did not fight the bracket. They just see a slightly soft edge on the left side of the field, and they say so. Separating the adjuster from the judge is the cheapest quality system we know of: it costs one plane ticket and removes an entire category of self-deception from the machine's baseline.

We institutionalized the same principle in our lighting work - the image gets argued over on a bench, by more than one person, before anything is welded. A vision system's baseline image is set exactly once, on the day of installation, under commissioning pressure. Every future drift, every retrain, every "the camera has been flaky lately" gets compared against that baseline. It deserves a witness.

The practical list

There are also the unglamorous reasons, which any integrator's insurance carrier will recite for free:

• Cameras live in inconvenient places. Somebody is on a ladder or inside a guarded cell, and lone-worker rules exist because of what has happened when they were ignored.

• Lockout coordination. The person at the camera needs the line stopped; the person at the HMI needs it running to see product. Two people can sequence this safely. One person in a hurry improvises, and improvisation around interlocks is how incident reports get written.

• Live product tests need a hand at each end - one to feed parts, one to watch results. A solo installer either recruits an untrained operator or tests with the three parts that happen to be in arm's reach, which is how a system gets commissioned on a sample size of three.

• Radios and phone cameras help, and we use them. They fix the walking. They do not fix the judging - a photo of a monitor, judged by the person who made the adjustment, is still one set of eyes grading its own work.

What it looks like when it works

The two-person install ends with something the solo install almost never produces: a saved reference image that two people argued about and then agreed on, focus and aperture rings marked with witness paint, and a note in the machine file about why the aim is what it is. Years later, when someone bumps the bracket or swaps a lens at 2am, that is the recoverable state - the difference between a fifteen-minute fix and a week of "the vision system was never right after the maintenance shutdown."

The commissioning checklists we run on installs - camera, lighting, and the sign-off sequence between the two people - live in the Saber knowledge base. And if you have a vision system that was commissioned by one tired person and it has never quite been trusted since, that is a service call we know well. Tell us through the contact form.