Do occupancy sensors use cameras? Some do and most do not. The category spans at least six distinct technologies, only one of which involves an optical lens, and even that one does not have to store video or recognise faces. If you are screening sensors for a building and the camera question is what keeps you up at night, this is the explainer that maps the whole landscape before you shortlist anything.
The confusion is understandable. "Occupancy sensor" is an umbrella term, and the vendors underneath it use wildly different physics to answer the same question: is anyone here, and how many. Below we break down each type, where it is honest about its limits, and where Occivar™ sits.
01 · DO OCCUPANCY SENSORS USE CAMERAS?Do occupancy sensors use cameras?
Most occupancy sensors do not use cameras. The common types passive infrared (PIR), microwave, ultrasonic and the wall switches in your office detect heat or motion, not images. Camera-based, or optical, sensors are a separate, more capable category used where you need an accurate headcount rather than a binary occupied/empty signal.
The important distinction is between a sensor that has a lens and a system that keeps the footage. An optical sensor that runs its model on-device and discards every frame after counting is, from a privacy standpoint, closer to a thermal sensor than to a CCTV camera. Occivar™ is in that group: the optics never leave the device, no video is stored, and no face is ever matched. We cover the why of that design in privacy by design.
02 · WHAT ARE THE DIFFERENT TYPES OF OCCUPANCY SENSORS?What are the different types of occupancy sensors?
There are six broad families, ordered roughly from cheapest and crudest to most accurate.
| Sensor type | What it senses | Counts people? | Stationary person? | Privacy profile |
|---|---|---|---|---|
| PIR (passive infrared) | Changes in body heat | No (presence only) | No | Very high |
| Ultrasonic / microwave | Reflected sound or radio waves | No (presence only) | Partly | Very high |
| Thermal array | Coarse heat map | Approximate | Yes | High |
| Radar / mmWave | Reflected radio at millimetre wavelengths | Approximate | Yes | High |
| Time-of-flight (ToF) | Depth via light pulse timing | Yes | Yes | High |
| Optical / camera | Image frames + computer vision | Yes (accurate) | Yes | Depends entirely on architecture |
PIR is what sits behind the light switch that turns off when you stop moving it is a presence detector, not a counter. Thermal, radar and ToF give you a count without an image. Optical gives you the most accurate count and the richest data, and is the only family where privacy is an architecture choice rather than a property of the physics.
03 · WHAT IS THE DIFFERENCE BETWEEN AN OCCUPANCY SENSOR AND A MOTION SENSOR?What is the difference between an occupancy sensor and a motion sensor?
A motion sensor reports whether something moved; an occupancy sensor reports whether a space is occupied. The gap matters most when people are still.
A motion sensor (a basic PIR) goes quiet the moment you stop moving sit reading at a desk for ten minutes and it will declare the room empty and cut the lights. That is fine for a corridor and wrong for a meeting room. A true occupancy sensor is built to hold the "occupied" state through stillness, either by combining motion with a timeout or by using technology (thermal, radar, ToF, optical) that detects a present body directly rather than inferring it from movement.
04 · CAN AN OCCUPANCY SENSOR DETECT A STATIONARY OR MOTIONLESS PERSON?Can an occupancy sensor detect a stationary or motionless person?
It depends on the sensor type. A plain PIR motion sensor cannot a motionless person becomes invisible to it within seconds. Thermal, radar/mmWave, time-of-flight and optical sensors can, because they detect the presence of a body rather than its movement.
This is the single most common failure in cheap occupancy systems: the "ghost vacancy", where a room full of people sitting still reads as empty and the building reclaims the space or kills the climate control. We have written a full worked example of that problem in the ghost meeting room. If your use case involves seated, stationary people which is most offices, libraries and meeting rooms a motion-only sensor is the wrong tool.
05 · WHAT IS THE DIFFERENCE BETWEEN PIR AND MMWAVE OCCUPANCY SENSORS?What is the difference between PIR and mmWave occupancy sensors?
PIR detects changes in infrared heat and only sees movement; mmWave radar emits millimetre-wavelength radio and can detect the micro-movements of a stationary person, including breathing. mmWave is more sensitive and can give a rough count; PIR is cheaper, lower-power and presence-only.
In practice:
- PIR is binary, blind to stillness, and trivially cheap. Good for lighting control in transient spaces. Useless for counting.
- mmWave holds presence through stillness and can estimate how many people are in a zone, but its count degrades as density rises and it struggles to separate people standing close together. It is genuinely camera-free, which makes it attractive on privacy grounds.
mmWave is a real step up from PIR, but "approximate count" is the ceiling. When you need to know that the third floor hit 84% of capacity at 11am, not just that "several people are present", you are into ToF or optical territory.
06 · DOES A CAMERA-BASED OCCUPANCY SENSOR STORE VIDEO OR RECOGNISE FACES?Does a camera-based occupancy sensor store video or recognise faces?
Not necessarily, and a well-designed one does neither. Whether a camera sensor stores video or runs facial recognition is a decision the vendor makes in software it is not forced by the hardware. The privacy risk lives in the architecture, not the lens.
Occivar™'s optical sensors process every frame on the device, extract an anonymous count, and discard the image. No footage is written to disk, nothing is streamed to a server, and the model has no facial-recognition capability to begin with. There is nothing to match a face against, by design. We hold that line with no exceptions, which is the whole point of no biometrics, no demographics, no exceptions.
The privacy risk of a camera sensor is not the lens. It is whether the footage ever leaves the device. On Occivar™, it never does.
07 · HOW DOES EDGE-AI INFERENCE MAKE A CAMERA SENSOR PRIVACY-SAFE?How does edge-AI inference make a camera sensor privacy-safe?
Edge AI means the computer-vision model runs on the sensor itself, so raw images are turned into a number before anything is transmitted or stored. The pixels never leave the box; only the count does.
This is the mechanism that lets an optical sensor be as privacy-respecting as a thermal one while keeping optical accuracy. A cloud-based camera system, by contrast, has to send footage somewhere to process it and that footage is the liability, the breach surface and the thing a procurement reviewer will rightly object to. On-device inference removes the footage from the equation entirely. The full design rationale is in edge AI, and the data flow in our security model.
08 · HOW ACCURATE ARE THE DIFFERENT OCCUPANCY SENSOR TYPES?How accurate are the different occupancy sensor types?
Accuracy tracks capability: presence-only sensors are reliable at the binary question and incapable of counting, while optical sensors are the most accurate counters when honestly calibrated.
Accuracy claims deserve scepticism, including ours. A headline figure like "98% accurate" is meaningless without knowing where the errors cluster doorways, tailgating, extreme density. We publish our failure modes rather than hide behind a single number, which we argue for in counting accuracy: ground truth and humility. The right question to ask any vendor is not "how accurate are you?" but "show me your error distribution and how you calibrate."
09 · WHICH OCCUPANCY SENSOR IS RIGHT FOR OFFICES AND CAMPUSES?Which occupancy sensor is right for offices and campuses?
For lighting and HVAC triggers in corridors and bathrooms, PIR is fine and cheap. For accurate space utilisation across desks, meeting rooms and floors the data that justifies a lease decision or a fit-out you want a counting sensor that holds presence through stillness and is privacy-safe by architecture.
That is the gap Occivar™ fills: optical accuracy with edge-AI inference, so you get a real headcount with no video stored, no faces, no biometrics. The counts feed straight into occupancy analytics for the trends that drive workplace and campus decisions. If you are mapping the sensor landscape for an office or a university campus, the camera question is the right one to start with and the answer can be "yes, optical, and nothing is ever stored."
Want to see anonymous, camera-free-by-design counting against your own floor plan? Book a demo and we will walk through the data flow end to end.