Overcapacity Detection: Crowd & Occupancy Limit Alerts
Every crowded space has a number it must not pass: the occupancy the fire code assumes, the load the exits can actually clear. Overcapacity is not a cosmetic problem. Above that number, an evacuation stops being a drill and becomes a crush risk, and the manager usually finds out from a headcount at the door that stopped matching reality an hour ago.
Overcapacity detection closes that gap with the cameras you already have. It counts people in defined zones on each stream, compares the count against a threshold you set, and alerts while the number is still a problem you can fix – before it becomes an incident report.
How the Counting Works
The rule runs person detection on each frame – the same object detection foundation as other behaviour rules – and tracks entries and exits against a zone. You define the zone to match the real constraint: one retail floor, one mezzanine, one assembly hall. Three parameters shape the alert. Threshold is the maximum count for the zone. Dwell requires the count to stay above the threshold for some seconds, which filters out the momentary spike when a tour group walks through. Direction lets you count a queue in one lane without counting the people browsing next to it.
Because counting happens on the stream itself, no new camera infrastructure is required – the appliance takes ONVIF and RTSP feeds from equipment you already own and processes them on-site. Multiple zones can run on one camera, with separate thresholds, and every alert is logged with the count and timestamp so occupancy history builds up for free.
From Alert to Action
An alert is only useful if something happens next, so the rule is designed to plug into the response you already have. The screen shows the zone, the current count and the trend; that is what a supervisor needs to open a second exit or slow the inflow. Where a gate or turnstile system exposes an interface, the same alert can trigger flow limiting automatically, so ingress stops before staff even reach the floor. Occupancy loads and safe egress are regulated for good reason – OSHA and local fire codes assume you know your number.
Where It Pays Off
Retail floors are the obvious case, and the same camera estate often runs queue and loss-prevention rules – see our work on AI cameras for retail stores and the wider retail industry page. Transport hubs and stations face the same physics with higher stakes. Schools and universities run it during assemblies and open days, where one overflowed hall is a real evacuation problem – the campus layout is covered in our school campus safety guide. Event venues round out the list: the number at the gate is not the number in the hall.
Honest Limits
Dense crowds push every counting method. When bodies occlude bodies at festival density, counts drift low; the rule is most reliable up to the densities typical of retail, transit and school spaces. Camera angle matters too – a steep overhead view counts far better than a shallow one across a room. And the rule counts, it does not predict: it tells you the number is too high, not which individual will cause trouble.
One more thing worth knowing before you buy: counting accuracy is testable before commitment. During evaluation we run the rule against recordings of your actual space, at your actual peak times, and show you the error rate on that footage. A vendor who cannot demonstrate that on your video is asking you to find out after the invoice.
It Works Alongside the Rest of the Library
Overcapacity pairs naturally with crowd gathering detection, which flags the cluster forming before the room is full, and with loitering detection at entrances. When something does go wrong, fall detection and fall protection detection cut response time, and for care settings there is dedicated work on elderly fall detection. After hours, the same cameras switch to intrusion detection.
See how it runs on a real site in smart city surveillance.
See it applied end to end in our guide to counting people with the cameras you already have.
Frequently Asked Questions
How accurate is the counting?
In normal retail, transit and school densities, counts stay reliable when the camera has a reasonably steep view. Extremely dense crowds cause undercounting everywhere; we test against your real footage during evaluation and tell you what to expect.
Can it limit entry automatically?
Yes, where your gate or turnstile system exposes an interface the alert can trigger – otherwise it routes to staff screens and phones for manual flow control.
Do we need new cameras?
No. The rule runs on existing ONVIF and RTSP streams. If a particular camera’s angle makes counting unreliable, we will tell you before you spend anything.
What happens outside opening hours?
The same cameras can switch to other rules – intrusion, loitering, tampering – so the estate stays useful around the clock.
Next Step
Send us the floor plan and the occupancy limit that applies. From the 198+ algorithm library this is configuration, not development – the first demo takes about 7 days where the detection already exists. Start with a Starter Kit or tell us your scenario.