Most people assume that counting people means watching them, and watching means a camera. It is the obvious mental model: point a lens at a doorway, run some software, get a number. So the idea of counting accurately with no camera at all sounds like either a trick or a compromise, a method that must be giving something up to avoid the video feed.
It is neither. Camera-free counting is not a workaround for the camera; it is a different way of measuring that happens to sidestep the image entirely. Instead of recording what a person looks like, it measures the physical fact that a body of a certain size and shape moved through a certain space at a certain time. That is enough to count, to time, and to trace a path, and it turns out you do not need a picture to do any of it. This post explains the mechanism: how the sensing actually works, what each method captures, and, just as important, what it never captures.
This is a technology explainer rather than a comparison. If you have already decided the camera is a problem and want the case for switching away from camera counters, that argument lives in its own post. Here the goal is to explain how the no-camera methods work in the first place.
How do you count people without a camera?
You count people without a camera by measuring the physical shape and movement of bodies instead of recording their images. Time-of-Flight depth sensing fires infrared pulses at an entrance and measures how long they take to return, building a depth map that registers a person as a moving shape and counts them as they cross, capturing geometry rather than any picture. Signal sensing detects the radio signals a phone emits to follow movement through an interior without storing a MAC address. Neither method produces video or a recognisable face, so there is no image to store, blur, or leak, while still delivering accurate counts, dwell times, and flow paths.
Time-of-Flight depth sensing: measuring shape, not appearance
The first method counts by measuring distance. A Time-of-Flight sensor mounted above an entrance fires rapid, invisible infrared pulses downward and measures precisely how long each pulse takes to bounce back. Light travels at a known, constant speed, so the return time converts directly into a distance, and doing this across a grid of points produces a depth map: a live, three-dimensional read of how far away every surface in view is, refreshed many times a second.
When a person walks under the sensor, the floor that was a fixed distance below suddenly has a raised object at head and shoulder height moving across it. The depth map registers that object as a shape of a characteristic size and profile, tracks it as it moves through the field of view, and increments the count as it crosses the line. What the sensor is reading is geometry, the height and outline and movement of a form, not appearance. It has no idea what colour the person is wearing, what their face looks like, or who they are, because none of that information is in a depth map. A depth map is a field of distances, and distances do not carry identity.
That is the crucial difference from a camera. A camera records the light reflecting off a scene, which is exactly the information that makes a face recognisable. A Time-of-Flight sensor records how far away things are, which does not. The output looks nothing like a photograph; it looks like a topographic read of the space, and a person appears in it as a bump moving across the floor. This is also what makes it device-independent: it counts the body itself, so it does not matter whether the person carries a phone, and it does not depend on any signal they emit. For the sensor itself in more depth, see how Time-of-Flight sensing works, and for how it stacks up against image-based approaches, computer vision vs Time-of-Flight covers the trade-offs.
Signal sensing: following movement through an interior
Depth sensing is excellent at a defined crossing point, an entrance, a corridor, a choke point where everyone passes through a bounded field of view. Following how people move through a larger interior, across a whole floor of a store or a terminal, is a different problem, and the second method addresses it by detecting a signal the person's device is already emitting.
A phone that is switched on is a small radio, and it emits signals as part of staying connected, enough for sensors distributed through a space to detect its presence and estimate where it is by comparing the signal across receivers. That estimate, updated as the person moves, traces a path through the interior: entered here, lingered there, moved to this zone, left by that door. It works even when a phone is in airplane mode, because a device still emits detectable signals in that state, which is why the method does not depend on the visitor connecting to anything.
The point that matters for this explainer is what the method does not keep. A phone's signalling can include a hardware address, the MAC address, which is a device identifier. A signal-sensing method that stored it would be capturing something many regulators treat as personal data. The camera-free approach described here stores no MAC address by default; it uses the signal to locate a moving point in space and to trace its path, not to record which device it was. The output is a trajectory, a line through the building over time, with no identity attached to it. It knows a phone-carrying person moved this way; it does not know whose phone.
Where the data goes
Neither method does its real work at the sensor. This is a detail worth being precise about, because it is often described wrongly. The sensor's job is to produce the two identifier-free feeds: the depth read at the crossing points and the signal read across the interior. Neither feed on its own is the full picture. The depth sensor knows a body crossed a line but not where it went afterwards; the signal read knows a point moved through the interior but is not the authoritative entrance count.
Ariadne measures this with Hybrid Fusion, its patented camera-free method. Time-of-Flight depth sensing counts every visitor at the entrances, capturing geometry rather than images, while patented phone signal sensing follows movement through the interior, detecting the signals a phone emits even in airplane mode, and tracks that movement to about one-metre precision. The sensor streams both feeds to Ariadne, where Hybrid Fusion combines them into one trajectory per visit and computes counts, dwell, and paths. The streams carry no identifier: no MAC address, no device ID, no biometric data, and no camera is involved. Identifiers are stored only when a visitor explicitly opts in, which keeps the method GDPR-friendly and outside biometric territory.
The word to hold onto is *centrally*. The fusion that stitches the entrance count and the interior movement into a single visit happens in the Ariadne platform, not on the sensor and not at the edge. The sensor is a producer of raw, identifier-free feeds; the intelligence that turns those feeds into counts, dwell times, and paths runs after the data leaves the device. That is why the method can combine the accuracy of a depth count at the door with the reach of signal sensing across a floor: the two are reconciled where the full context is available, not in isolation on a single unit.
What no-camera counting captures, and what it never captures
It is worth stating the boundary plainly, because the whole method is defined by it.
What it captures:
- Counts: how many people entered and left, at each entrance, over any period.
- Dwell: how long people stayed, in the space overall and in specific zones.
- Paths: how people moved through the interior, which routes are busy, where they linger, where they turn back.
What it never captures:
- Images or video of any kind, because no camera is involved.
- Faces, and therefore no facial geometry and no basis for facial recognition.
- Biometric data, meaning no physical characteristic used to identify a person.
- A stored device identifier by default, because the MAC address is not retained.
That boundary is the reason the method carries so little privacy risk. There is no footage that could be subpoenaed, leaked, or repurposed, because none is recorded. There is no biometric template that a breach could expose, because none is created. The measurement is of movement in aggregate, and the identity of the people moving is not just protected, it is never in the system to begin with. That is a stronger position than a camera method that captures faces and then promises to handle them carefully, because the risk it removes is the risk of the data existing at all. It is also why this approach lets a venue measure audiences without facial recognition, where a camera-based count would raise exactly the questions the no-camera method never has to answer. For the category term and how "anonymous" is defined precisely here, see anonymous people counting.
Where camera-free counting fits
The mechanism suits the places counting actually matters. At an entrance, the depth method gives an accurate, device-independent count of everyone who crosses, which is the foundational number a retailer, a venue, or an airport builds on. At a choke point, a corridor, a gate, an escalator, the same crossing-based count measures flow through a bounded space. Across a larger interior, signal sensing traces where the crowd goes after it enters, which is the difference between knowing how many came and knowing what they did.
The privacy payoff is not a side benefit; for many operators it is the deciding factor. A camera at an entrance invites questions about footage retention, staff and visitor consent, and, increasingly, biometric regulation. A method that records shapes and paths and no images sidesteps that whole category of concern, while still producing the counts, dwell, and flow data an operator needs to run the space. That combination, useful data without the video, is why camera-free counting has moved from a niche privacy choice to a mainstream one. If you want to see it applied end to end, camera-free people counting walks through the full picture.
FAQ
How do you count people without a camera?
By measuring the physical shape and movement of bodies rather than recording their images. Time-of-Flight depth sensing at an entrance measures the distance to a moving form and counts it as it crosses, capturing geometry not appearance, while signal sensing detects the radio a phone emits to trace movement through an interior without storing a device identifier.
Is no-camera counting less accurate than a camera?
No. Depth sensing counts the body directly and is not affected by lighting or by what someone is wearing, and it is device-independent because it does not rely on a phone. It produces counts, dwell times, and paths comparable to what an operator expects, without recording any image.
Do I need cameras?
No. Ariadne counts with Hybrid Fusion: Time-of-Flight depth sensing plus patented phone signal sensing, never cameras. Time-of-Flight captures geometry rather than images, and signal sensing captures no MAC address by default, so the measurement involves no video, no faces, and no biometric data.
Does signal sensing track my specific phone?
Not by identity. It detects that a phone-carrying person is moving through a space and estimates the position to trace a path, but it stores no MAC address by default, so the output is a trajectory through the building with no device or person attached to it.
Where does the counting actually happen?
Centrally, not on the sensor. The sensor streams two identifier-free feeds, the depth read at crossing points and the signal read across the interior, to the Ariadne platform, where they are combined into one trajectory per visit and turned into counts, dwell, and paths. The fusion happens after the data leaves the device, not at the edge.
