GTU Technopark · Gebze · engineered in TürkiyeGDPR/KVKK · data processed anonymouslyinfo@cxteknoloji.com
EN
Book a 15-min demo
GVN-02 · Zone & Perimeter Security

Fire & Smoke Detection

Detect flame and smoke from video within seconds, without waiting for a sensor.

Visual
Event frame and camera location in every alert
<10 s
From detection to alert
0
Additional sensors required

A smoke detector waits for smoke to reach it. A camera sees the moment smoke forms — in high-ceilinged, open or windy spaces the difference is measured in minutes.

The model evaluates flame colour, motion texture and the spreading geometry of smoke together; steam, dust and vehicle exhaust are filtered out. A visual alert is sent on detection, so the recipient does not have to walk to the site to confirm the event.

How it works

01 / COLOUR + TEXTURE

Searching for the flame signature

Colour histogram and flicker texture are evaluated together.

02 / SPREAD

Smoke geometry

The rising, widening motion pattern is separated from steam.

03 / ESCALATION

Visual alert

First to security; if density increases, automatically up the chain.

Early detection changes the nature of the damage

In a fire, cost grows exponentially rather than linearly with time. The minute of first response determines whether this is an extinguishable incident or a production stoppage. The metric that matters is therefore not alarm count but detection-to-response delay.

Sample dashboard
Delay by detection method · seconds
Video analytics8 s
Smoke detector (enclosed space)55 s
Smoke detector (high ceiling)150 s
Human report120 s

In high-ceilinged warehouses it takes minutes for smoke to reach a detector; a camera removes that delay entirely.

Monthly false alarm count · false alarms −72%
Jan18
Feb14
Mar11
Apr9
May7
Jun6
Jul5
Aug5

Sources such as steam, welding sparks and vehicle headlights are eliminated during the pilot; thresholds are updated until the rate stabilizes.

Use cases

Warehouse

High-rack area

Volumes where detectors fall short.

Recycling

Open-yard piles

Spontaneous combustion risk.

Production

Welding and cutting

Incipient fires started by sparks.

Parking

Enclosed car park

The first seconds of a vehicle fire.

Technical requirements

CameraStandard visible-light camera; thermal optional
RangeFlame source ≥ 20 px in frame; effective up to 60 m
NightFlame detection achieves higher accuracy at night
IntegrationFire alarm panel, IoT relay, siren, mobile push
VerificationAlerts include event footage and camera location

Frequently asked questions

How does Fire & Smoke Detection work?

A smoke detector waits for smoke to reach it. A camera sees the moment smoke forms — in high-ceilinged, open or windy spaces the difference is measured in minutes.

Does it replace our fire alarm system?

No. The fire detection system installed as required by regulations and insurance stays in place; the camera adds an early-warning layer with footage in front of it. A camera only detects what it sees: it cannot see a fire that starts inside a suspended ceiling, in a cable duct or in an electrical panel and has not yet produced visible smoke.

Are our existing cameras suitable for this solution?

Required camera specification: Standard visible-light camera; thermal optional. Before installation we assess your camera inventory free of charge and report which points are suitable and, if needed, which camera should be added.

How accurate is it in the field?

We do not publish a single general figure. Results vary with the camera angle, the size of the flame in the frame, lighting conditions and misleading sources on site. Thresholds are calibrated on your own site during the pilot, and detection performance is measured there.

Which systems does it integrate with?

Fire alarm panel, IoT relay, siren, mobile push. Events are published over REST API and MQTT, so it connects to your existing MES, ERP, VMS and alarm infrastructure.

Is it compliant with KVKK, Türkiye's data-protection law?

By default, video is processed on an on-premise server inside the facility and raw video is not sent to the cloud; if preferred, processing can also run in the cloud. Facial recognition is not used; analytics outputs are anonymous and event clips are deleted automatically when the defined retention period ends.

See this solution on your own site
Let's test it with your existing camera in a 15-minute demo call.
Book a 15-min demo

Other solutions in Zone & Perimeter Security

GVN-01 · 93%
Zone Violation Detection
Entry, dwell-time and headcount rules for every polygon you draw.
GVN-03 · Perimeter
Perimeter Security
Monitor unauthorized entry, vehicle approach and camera tampering.
GVN-04 · ADR
Warehouse & Hazmat Monitoring
Enforce flammable-goods zones, stacking rules and escape routes.
GVN-05 · 99%
Licence Plate Recognition (ANPR)
Automatically record the plate, time and image of every vehicle entering your facility.

Technical articles on this topic

Camera-Based Fire Detection
IN TURKISH
Tersanede İSG için Yapay Zekâ
IN TURKISH
OSB Güvenliği
IN TURKISH
Gaz Kaçağı Kamerayla Tespit Edilebilir mi? Dedektör, OGI ve Görüntü Analitiği
IN TURKISH
Tersanede Sıcak İş ve Yangın
IN TURKISH
Kocaeli, Gebze, Tuzla ve Anadolu Yakası Sanayisi İçin Görüntü Analitiği
IN TURKISH
Alan İhlali Tespiti
IN TURKISH
Akıllı Şehir Analitiği