Recording is not protecting
In most facilities the camera system is a recording system: you go back and look after something has happened. Watching the footage after a theft does not bring the stolen goods back. What you actually need is to know while it is happening.
The classic answer is motion sensors and motion-triggered camera alarms. In practice these get switched off within weeks, because:
- Branches in the wind, rain, fog and moving shadows raise alarms.
- Cats, dogs, birds and insects keep the system firing all night.
- Headlights, passing clouds and automatic lighting trigger it.
The result is a security team that receives 40 notifications overnight and mutes the system the next morning. A muted alarm system is equivalent to no system at all.
The difference: person detection, not motion
AI Kameram looks at what is in the image. The model determines whether the object in the frame is a person. That single difference cuts false alarms dramatically: a swaying branch is not a person, a cat is not a person, a shadow is not a person.
Vehicle detection can likewise be configured as a separate rule. A vehicle entering the site outside working hours can raise an alarm, while parked vehicles are classified as stationary objects and do not alarm over and over.
Zone and time rules
The second pillar is rule definition. In the panel, for each camera you:
- Draw the monitored zone on the image. Only the fence line, the doorway or the storage yard can be monitored, leaving the neighbouring road out of scope.
- Define time ranges. Seeing a person during working hours is normal and raises no alarm; the same image outside those hours is an alarm.
- Assign different rules to different zones. A loading dock can be open 08:00–18:00 on weekdays and fully closed on Saturdays.
This flexibility lets the system follow the real working rhythm of the facility, so normal staff activity never raises an alarm.
A practical example: A warehouse monitors its main entrance, loading dock and rear fence line. On weekdays between 07:00 and 19:00 only the fence line is active — people on the dock are expected. After 19:00 all three zones go live. At 02:14 a person is detected on the fence line: the guardhouse device sounds its siren, the security supervisor receives an automated call, and an e-mail with a screenshot arrives.
Extra layers that cut false alarms
Static object separation
Person-like objects standing still in the scene (a mannequin, a poster, a hi-vis jacket on a hook, a fixed shadow) do not alarm repeatedly; the system separates out detections that never move.
Headlight glare and vehicle overlap
Glare from vehicle headlights at night and detections overlapping a vehicle are evaluated separately, so vehicle movement is not read as a person entering.
Second-stage verification
For critical events the detection passes through a second visual verification before becoming an alarm, so ambiguous frames do not generate needless alerts.
What happens when an alarm fires?
- Guardhouse device — siren sounds, the intruder icon flashes red, the LCD shows the location.
- Automated phone call — registered contacts are called.
- SMS and e-mail — the e-mail includes a screenshot of the moment.
- Mobile push — instant notification to panel users.
- The event video and evidence frame are archived.
The officer on duty does not merely hear the alarm — they see it, and can decide within seconds whether a response is genuinely required.
Where does it help most?
- Warehouses and logistics sites — large areas left empty at night
- Construction sites — risk of material, cable and fuel theft
- Industrial plants — scrap, metal and equipment theft
- Car parks — vehicle and in-vehicle theft
- Farms and agricultural businesses — remote locations with long perimeters
- Residential and facility management — perimeter security and access control
See the use cases page for the full list.
Frequently asked
How is it different from a motion sensor?
A motion sensor treats every movement as an alarm: swaying branches, cats, rain, shadows and insects included. AI Kameram checks whether there is genuinely a person in the image, which eliminates the empty alarms that ring all night.
Will staff working during the day trigger alarms?
No. Monitoring hours are defined per camera and per zone. Person detection during working hours raises no alarm; alarms occur only outside the defined hours or inside restricted zones.
Can it detect people in the dark?
Yes. The camera's IR night vision is sufficient; the system performs person detection on black-and-white night footage using night-specific filters.
Can it detect vehicles as well?
Yes. Vehicle detection can be configured as a separate rule alongside person detection — for example a vehicle entering the site outside working hours can raise an alarm. Parked, stationary vehicles are separated out so they do not alarm repeatedly.