Home / AI / Fire / Flame detection

Flame detection

Confirms flame in the image across consecutive frames, filtering out brief flashes and reflections to report fires early.

Request a demoUser manual

Flame detection

Flame detection examines the image in the area you choose at short intervals and confirms the presence of flame across consecutive frames. The flame must remain visible for the fire confirmation time you set; brief flashes such as sun glare, headlights or welding sparks are filtered out by this window. If the video is frozen, ProVife does not perform fire confirmation, so a stuck frame never produces an event. It provides early warning where spot detectors react too late, such as warehouses, open storage yards, parking garages, waste facilities and high-ceiling spaces. The event is reported with an image marking the area where the flame was seen, and it starts fire scenarios.

How it works

  1. Draw the area where flame should be monitored; it can be a storage area, a machine or a stock yard.

  2. Enter the fire confirmation time; the flame must remain visible throughout this period.

  3. Use the confidence threshold and the consecutive confirmation count to set how certain the event must be.

  4. ProVife examines the area across consecutive frames and skips confirmation on frozen video.

  5. Once confirmation is complete, an event is created and the fire scenario kicks in.

Settings

Fire confirmation timeHow long the flame must remain visible to trigger an event; filters out brief flashes.
Confidence thresholdThe minimum confidence required to accept a flame detection.
Consecutive confirmationHow many evaluations must repeat the flame detection before an event.
AreaThe region where flame is searched for.
SeverityCritical level is recommended for fire events.

What it produces

  • Flame detection event: confidence value, duration and the region where the flame was seen.
  • Event image marking the part where the flame is.
  • A label over the flame region in live video.
  • The event is listed in AI Playback under a fire heading.

What you can do with the event

  • Critical notification, pop-up window, audible alert and WhatsApp message with a ready-made scenario.
  • Running the siren for five minutes through a smart relay.
  • A “Fire alarm. Please evacuate the building.” announcement over the speakers.
  • Forwarding the event via MQTT to building automation and fire management systems.
  • Showing the event location on the floor plan.

Where it is used

Logistics

In high-ceiling warehouses, flames starting on upper rack levels are spotted early.

Waste and recycling

Piles prone to spontaneous combustion in open storage yards are monitored through the night.

Parking lot

A fire starting in a vehicle in an indoor parking garage is sent to security instantly.

Agriculture

Flame around haystacks and barns is reported to the farm owner.

Camera placement

  • Position the camera so it directly sees the spot where a fire could start.
  • Avoid angles facing the sun, glass and shiny metal surfaces.
  • Keep places where flame is normal, such as welding, stoves or fireplaces, out of the area.
  • Rather than making the area very large, monitor risky zones with separate scenarios.

Related analytics

Try flame detection on your own camera.

Request a demo