Thermal and Depth Sensors

Thermal arrays and depth sensors produce a coarse image of a space from which people can be counted without being recognisable. For headcount questions they are the usual answer.

The technical evidence in “Thermal and Depth Sensors” describes the building, while the effort required to install, test and maintain it is project work. When teams research time tracking with screenshots, the product page can provide time and project context for that operational effort without replacing the sensor platform as the source of truth for physical occupancy.

What they do

Thermal arrays read heat across a low-resolution grid and infer bodies from shape and temperature.

For a public, independent reference related to “Thermal and Depth Sensors”, consult the ASHRAE technical resources. Its principles provide a useful check on scope, terminology, governance and the claims made during procurement or review.

Depth sensors measure distance to surfaces and infer people from height profiles, usually mounted overhead.

Both produce a count rather than a binary, and neither captures an identifiable image — which is the main reason they are chosen over cameras.

What they are good at

Headcount in rooms: how many people in this meeting.

Entry and exit counting at a doorway, which gives dwell time.

Counting while people are still, which infrared cannot do.

And doing it without an image, which substantially simplifies the consultation and the data protection position.

The failure modes

Crowding. Two people close together can merge into one detection, particularly in thermal.

Height assumptions. Depth sensors tuned for standing adults miscount seated people, children, and wheelchair users — which is an accessibility problem as well as an accuracy one and has its own note.

Heat sources again, for thermal: a radiator, a sunlit floor, a printer.

Mounting height, which both are sensitive to and which retrofits frequently get wrong because the ceiling is where the ceiling is.

And doorway geometry: a wide entrance or a door that swings into the field produces persistent errors.

Calibration matters more here

Because the output is a number rather than a binary, errors are less obvious.

A room reporting four when six are present looks plausible and is wrong by a third.

Count manually against the sensor for a week in the busiest spaces. This is the single most valuable deployment task and it is routinely skipped.

The image question

Suppliers describe these as privacy-preserving, and compared with cameras they are.

Check what is actually transmitted and stored: some devices process locally and send counts, others send frames.

Ask directly, get it in writing, and treat it as a procurement requirement rather than a feature description.

Cost and where to spend

Substantially more per unit than infrared.

Which argues for mixing: thermal or depth in meeting rooms and at entrances where counts matter, infrared at desks where binary is enough.

A uniform deployment of the expensive option is how budgets get spent on resolution nobody needed.

What to check

Does your device send counts or frames, and is that in the contract?

Has anybody manually verified counts in your busiest room?

Is the mounting height within the stated range?

And how does it handle seated people and wheelchair users?