Proof of visitors
Museums, libraries and public bodies report visitor numbers to funders and councils.
People counting: how it works and what the data is good for
What is people counting, how does people counting work, and which sensor methods exist? A plain overview of the people counting technology, what each method records and where the line runs between a count and personal data.
What people counting measures
People counting — also called footfall counting — measures how many people pass a point, usually an entrance, and in which direction. From these crossings come three kinds of figures:
- Visitor numbers: how many came in per hour, day, month or year.
- Occupancy: entries minus exits, the number of people inside right now.
- Patterns: peak hours, busy weekdays, seasonal trends, the effect of an event or a change in opening times.
None of these figures needs to know who someone is. A count is a number, not a record about a person.
Sensor methods compared: infrared beam, thermal, 3D stereo, time of flight, camera with edge AI
The people counting sensor methods differ in what they see, where they work well and what they record.
| Method | How it counts | Works well | Limits |
|---|---|---|---|
| Infrared beam | A light beam across the door is interrupted | Narrow doors, little traffic, low budget | Groups side by side count as one; direction needs two beams |
| Thermal sensor | Detects body heat from above | Sites that rule out optical sensors | Low resolution; affected by warm surroundings |
| 3D stereo | Two lenses measure height and separate people | Dense crowds, busy entrances | Dedicated sensor per door; limited mounting heights |
| Time of flight | Measures distance with reflected light | Dark areas, overhead mounting | Limited range and coverage per sensor |
| Camera with edge AI | A detection model in the camera finds people and counts line crossings | Wide entrances, existing cameras, vehicles too | Needs suitable light and mounting position |
Wi-Fi and Bluetooth tracking is sometimes sold as counting, but it counts devices rather than people and works with device identifiers — which are personal data. Terms such as counting line, occupancy or edge processing are defined in the people counting glossary.
Where the counting line belongs
Whatever the sensor, the virtual line where crossings are counted decides the quality of the figures. It belongs where everybody has to pass and nobody stops: just inside the door, across the full width, at right angles to the walking direction. A line in front of a reception desk or a ticket queue counts people who hesitate, turn and cross again.
The sensor needs to see people some steps before and after the line, so it can tell the direction.
What is personal data and what is not
A figure such as “318 visitors on Tuesday” is not personal data — no one can be identified from it. An image in which people can be recognised is personal data, and so is a device identifier that is followed from place to place.
The decisive question is therefore not whether a camera is involved, but where the image is processed and what leaves that place. Architectures are compared in on-camera counting compared to server-based video analytics; the legal side is covered in people counting and the GDPR, and the case of sites without any outbound connection in people counting without cloud, server or internet.
Typical use cases
Capacity limits
Venues and public buildings keep to a maximum occupancy — see occupancy monitoring with a live capacity limit.
Staffing and opening hours
Peak hours show when more staff are needed and which opening times are actually used.
Vehicles
The same principle counts cars in and out of a car park and derives the free spaces.
What a system costs to own
The purchase is only part of the cost. Over five or ten years, other items often weigh more:
- Sensors: a dedicated device per door, or an app on cameras that are already there.
- Licence model: a one-time licence or a subscription that runs as long as you count.
- Infrastructure: a server, a gateway or a cloud platform — or none of these.
- Installation and maintenance: mounting, cabling, updates and replacement over the years.
How to choose
Five questions narrow it down quickly:
- How wide is the entrance, and how many people pass at the busiest moment?
- Do you need occupancy, or only visitor numbers?
- May data leave the building, and may the device connect to the internet?
- Are there cameras at the entrances already, and which ones?
- Who will look after the system in five years?
If you want to see how one camera-based approach answers these questions, the Vicodis products are described separately.
FAQ
Questions about people counting
01
What is people counting?
People counting is the automatic measurement of how many people pass a point or are inside an area. A sensor at the entrance registers each crossing and its direction, and software turns the crossings into visitor numbers, occupancy and trends over time. Good systems record counts only, not who the people were.
02
How does people counting work with a camera?
The camera looks down on the entrance and a detection model finds people in each image. When a detected person crosses a virtual line, the crossing is counted in its direction. With edge AI this happens inside the camera, so only the count leaves the device and no image is stored.
03
Which sensor method counts best?
There is no single best method; it depends on the entrance. Infrared beams suit narrow doors with little traffic, thermal sensors suit sites that rule out cameras entirely, 3D stereo and time-of-flight sensors handle dense crowds well, and cameras with edge AI cover wide entrances and can reuse existing hardware.