What is loitering detection in video surveillance?
Loitering detection is an AI video capability that flags a person or vehicle that remains within a defined zone longer than a configured time threshold. It turns ordinary dwell into an early-warning signal for ATMs, building entrances, perimeters, car parks, and restricted areas, where lingering often precedes theft, vandalism, or trespass. Running as an analytics layer on a Video Management System, it tracks each object, times its presence in the zone, and raises an alert when the dwell limit is exceeded, with the clip bookmarked for review. VMukti provides loitering detection among its 26+ AI models, deployable at the edge for real-time alerts and fed into ICCC and multi-camera tracking workflows.
What loitering detection does
Loitering detection flags a person or vehicle that stays within a defined zone longer than a configured time threshold. It converts an ordinary behaviour — dwelling — into an early-warning signal, because in the right context lingering precedes theft, vandalism, trespass, or casing a target.
Where it is valuable
- ATMs and bank vestibules: someone waiting without transacting.
- Building entrances and lobbies: dwell outside hours.
- Perimeters and restricted areas: presence where no one should linger.
- Car parks: a person moving between vehicles, or a vehicle parked where it should not be.
How it works
The analytic detects and tracks each object in the scene, starts a timer when it enters the monitored zone, and raises an alert when the dwell time exceeds the threshold. Because thresholds and zones are configurable, the same camera can apply a strict rule to a sensitive doorway and a lenient one to a public area. On an alert, the clip is bookmarked so an operator can review what happened and decide whether to act.
Tuning for signal, not noise
The art is context. A bus stop or a queue is full of legitimate dwell; a fenced substation perimeter is not. Per-zone thresholds, time-of-day rules, direction logic, and a verification step keep loitering alerts meaningful. Combining loitering with object classification ensures the alert fires on a person or vehicle, not a swaying tree or a parked delivery.
From alert to response
Loitering rarely stands alone. When it fires, an operator often wants to follow the subject — which is where multi-camera tracking takes over, stitching the person's path across the camera network so security can intercept before an incident.
How VMukti delivers it
VMukti provides loitering detection among its 26+ AI models on a hardware-agnostic, ONVIF platform (1,000+ camera models), deployable at the edge for real-time alerts. Detections feed the Integrated Command and Control Centre (ICCC) and multi-camera tracking, so an operator can move from a dwell alert to following the subject across the estate under defined SOPs. The platform is STQC-certified and proven across 900+ deployments.
Related
Last reviewed: 2026-06-17