Which lens: how far a camera can actually identify someone

The four distances, and why they are not one number

EN 62676-4 defines four pixel densities, and each one buys a different answer. They are not marketing tiers. The distances below fall off with pixel density, so identification happens at roughly a tenth of the distance at which the same camera can tell that somebody is present.

Every camera we publish, by reach

Computed from each camera’s own horizontal field of view and sensor, not copied from a datasheet. Cameras are grouped by lens, because on a fixed sensor the lens is what changes the answer.

What the lens actually buys you

Across the published range the spread is narrow, and it is worth saying so plainly. The 2.8mm cameras identify between 4.4 and 4.8 metres. The 4.0mm cameras identify at 6.0 metres. Every camera here is 4MP, so the lens is doing all of the work: nothing in this range resolves further by having a better sensor.

That also sets the limit. If you need a face beyond about 6.0 metres, no lens in this range will do it, and the answer is a longer lens or a camera closer to the subject rather than a higher megapixel count on the same optic.

Choosing in the order that matters

Pick the body first. Where it mounts, whether it has to survive being hit, and whether it sits under its own infrared are decisions the lens does not touch. A vandal dome and a turret with the same lens resolve identically.

Then measure the distance from that position to the place a face has to be usable, and read it off the identify column. Quote from identify, never from detect. Specifying a camera on its detect figure and delivering detection when the customer expected a face is the most common way one of these jobs fails after the invoice is paid.

Common questions

What is the difference between detect range and identify range?

They are four different jobs at four pixel densities, set by EN 62676-4. Detect at 25 pixels per metre only tells you somebody is there. Identify at 250 needs ten times the pixel density, so it happens at roughly a tenth of the distance. A camera advertised at 60 metres of detect will identify a stranger at about 6.

Does a 4.0mm lens see further than a 2.8mm on the same camera?

Yes, and by a predictable amount. The sensor is unchanged, so narrowing the field spreads the same pixels across less scene, which raises pixel density on anything in it. The gain is width traded for distance, in both directions. The table above gives the measured figure for every camera we publish rather than a rule of thumb.

Why are your figures shorter than the manufacturer datasheet?

Because we compute them per camera rather than per lens. Manufacturers commonly publish one DORI figure for every camera sharing a focal length, which cannot distinguish two models with different lens angles. We work from each camera’s own field of view and sensor, which gives shorter numbers we can defend.

How do I choose between them?

Measure the distance from the camera position to the spot where a face has to be usable, then pick the lens that still identifies at that distance. Choose the body first for where it mounts and what it has to survive, because that decision is independent of the lens.