Video analytics: what works in practice and what is mostly marketing
KINNEX Team6 min read
Analytics turn cameras from recorders into sensors. Done well, they send an alert instead of leaving someone to watch a wall of screens. Done badly, they send false alarms until people mute them.
Reliable, common uses
- Intrusion and virtual fence: alert when a person or vehicle enters a defined area, especially at night.
- Line crossing and loitering: useful at gates and restricted zones.
- People counting and queue length: simple, valuable for retail, hospitality and facilities.
- ANPR: reading number plates at gates and car parks, with a database for allow and deny lists.
- Object left behind or removed: useful in lobbies and transit areas, with careful tuning.
- Safety-gear detection: helmets or vests in industrial areas.
Use with care
- Face recognition depends heavily on camera angle, lighting and image quality, and raises privacy and legal questions. Decide the purpose and the legal basis before turning it on.
- Behaviour detection can be impressive in demos and noisy in real life.
What makes analytics work
Good placement, adequate light, a clear definition of what counts as an alert, and tuning after installation. Every false alarm costs trust, so start with a few high-value scenarios rather than all of them.
Questions to ask
Where does processing happen, on the camera, on the recorder or in the cloud? What is the false-alarm rate in conditions like yours? Can you try it on your own site first?