People counting
People counting is the technology and practice of measuring the number of individuals passing through a specific point or area — typically a store entrance, corridor, or zone — using sensors, cameras, or AI-powered systems.
What is people counting?
People counting is the process of measuring the number of people entering, exiting, or occupying a physical space. It is the foundational technology behind footfall analytics and is used across retail, transport, hospitality, corporate offices, and public spaces.
While the concept is simple — counting people — the implementation varies enormously in sophistication, accuracy, and the additional insight it can deliver beyond basic counts.
People counting technologies compared
Infrared beam counters
The simplest and oldest technology. A beam of infrared light crosses the entrance; each time it’s broken, a count is registered.
Pros: Low cost, easy to install Cons: Cannot count groups walking abreast, counts objects as well as people, no directional data (entry vs. exit), no demographic or behavioural insight
Typical accuracy: 70–85%
Thermal imaging sensors
Ceiling-mounted sensors detect body heat signatures to count people passing underneath.
Pros: Works in all lighting conditions, directional (in vs. out) Cons: Expensive per entrance, struggles with very crowded environments, affected by ambient heat sources, no additional analytics
Typical accuracy: 85–95%
Stereo-vision / 3D depth cameras
Dual-lens cameras create a depth map to identify people by their 3D shape.
Pros: Good accuracy, directional, less affected by shadows than 2D Cons: Requires dedicated hardware at each entrance, limited field of view, no demographic or journey insight
Typical accuracy: 90–97%
Wi-Fi / Bluetooth tracking
Detects the signals emitted by mobile phones to estimate the number of people in an area.
Pros: Can estimate dwell time and repeat visits Cons: Requires people to have Wi-Fi or Bluetooth enabled, MAC address randomisation has severely reduced accuracy, significant privacy concerns, cannot detect people without smartphones
Typical accuracy: 40–60% (post MAC randomisation)
AI-powered computer vision
Uses deep learning models running on standard CCTV camera feeds to detect, count, and track individuals.
Pros: Works with existing cameras (no new hardware), highest accuracy, provides additional insights (demographics, heatmaps, journey mapping, queue analytics), can distinguish staff from customers Cons: Requires on-premise processing hardware (edge device), needs initial AI training per location
Typical accuracy: 95–99%
Beyond basic counts
The most significant advantage of modern people counting systems is that counting becomes just the starting point. An AI-powered system that can count people can also:
- Track journeys through the store to create path maps
- Measure dwell time in specific zones
- Generate heatmaps showing where people spend the most time
- Detect queues and measure wait times
- Estimate demographics anonymously
- Segment staff from customers for cleaner data
This transforms people counting from a basic operational metric into a comprehensive analytics platform.
Use cases by industry
Retail
The primary use case: understanding store traffic, measuring conversion, optimising staffing, and evaluating marketing campaigns.
Transport
Railway stations, airports, and bus stations use people counting for capacity management, crowd safety, and service planning. Aura Vision’s work with c2c railway is an example of computer-vision-based counting in transport.
Smart cities
Councils and urban planners use pedestrian counting on high streets and public spaces to measure economic activity, assess the impact of events, and plan infrastructure.
Corporate
Offices use people counting for meeting room utilisation, space planning, and occupancy management.
How Aura Vision counts people
Aura Vision uses AI-powered computer vision running on a compact edge device (the APU) connected to existing CCTV cameras. The deep learning models are trained in-situ — meaning the AI adapts to the specific cameras, angles, lighting, and environment of each location.
This approach means:
- No new cameras or sensors need to be purchased or installed
- Staff are automatically excluded from customer counts via uniform recognition
- Multiple entrances are consolidated into a single store total
- All processing is on-device — no video footage leaves the building
- Deployment takes the same day — truly plug and play
Frequently asked questions
What technologies are used for people counting?
Common technologies include infrared beam sensors, thermal imaging sensors, stereo-vision cameras, Wi-Fi/Bluetooth tracking, LiDAR, and AI-powered computer vision on standard CCTV cameras. Each has different trade-offs in accuracy, cost, privacy, and additional insight capabilities.
How accurate are people counters?
Accuracy varies by technology. Simple infrared beam counters typically achieve 70–85% accuracy, thermal sensors 85–95%, and modern AI-powered computer vision systems 95–99%. Accuracy is most challenged in crowded environments and with groups walking side-by-side.
Can people counters distinguish between staff and customers?
Most traditional people counters cannot. AI-powered computer vision systems can be trained to recognise staff uniforms and automatically exclude employees from customer counts, providing a more accurate picture of actual customer traffic.
Do people counters work with multiple entrances?
Yes, though the approach varies. Legacy sensors require a device at each entrance. AI-based systems can use multiple camera feeds to consolidate counts across all entrances into a single, deduplicated total for the store.