Object and anomaly detection
AI on live image and video streams, running onboard where bandwidth and latency demand it.
Intelligence
From sensor streams to decisions
Perimeters, energy sites and harbours generate more observation than anyone can watch. Intelligence is the layer that decides what deserves attention — and what should happen next.
Critical infrastructure
Persistent, auditable monitoring
The single flight is worthless
Infrastructure monitoring is not a photo shoot. Value comes from repeating the same observation until change becomes visible — and from being able to prove you did.- Persistent coverageRepeatable missions over perimeters, energy assets and harbour facilities, flown the same way each time so results are comparable.
- Auditable by defaultEvery flight leaves a record of where it went and what it saw, which is what turns monitoring into evidence.
- Change over timeThe comparison between this inspection and the last one is the product — not either inspection on its own.
Persistent monitoring
Capabilities
What the platform can recognise
Infrared and thermal
Thermal analysis for the conditions visible light cannot report — heat signatures, losses and overnight operations.
Fleet-side intelligence
The same analytical approach applied inward: log-derived health insight about the fleet doing the monitoring.
Many streams, one record
Every camera, many models, one picture
Video in, findings out
Drones, docks and the fixed CCTV cameras already on site all produce video nobody has time to watch. Autrik runs every stream through the models the site needs and hands back what was found, not more footage.- Any sourceLive video from drones and docks, and from fixed and PTZ cameras over RTSP or ONVIF, comes into one place, whoever made the camera.
- One record per findingDetections from several cameras and models are folded into one event: what, where, when, from which source and how sure, with the clip attached.
- Into AutrikOS or your platformEvents land on the live map and in alarms and reports in AutrikOS, or go with their metadata to your VMS, PSIM or own system through the API.
Many streams in, one record out
Model library
Models for the jobs sites actually have
Pick the models per site and per camera; add your own next to ours.
Number plates
Read plates at gates and on roads and log each one with time, place and the frame it came from.
Fire and smoke
Flames and smoke in visible or thermal video raise an alarm early, at landfills, depots and forest edges.
Crowd counting
How many people are in an area and how dense it gets, for events, stations and evacuations.
People and vehicles
Detect and classify people, cars, trucks and bikes, the base layer for perimeter and site security.
Vehicle tracking
Follow a vehicle from the gate camera to the drone and keep it under one track id across sources.
Marine vessels
Detect ships and small boats in harbours and along coasts, including craft that send no AIS.
Model library
From detection to action
Recognition is only half of it
What matters is the response a detection triggers, and how fast.
Alarm-driven dispatch
A detection raises an alarm, and an alarm can put an aircraft over the location automatically.
Into your security centre
Events flow into the systems your team already watches, instead of demanding another screen.
Documented outcome
What was detected, what was dispatched, and what was found — recorded as a chain rather than three separate systems.
Where this stands today
What ships, what is being built
What ships today: detections on live streams and recordings, the DJI on-board recognition, thermal maps with a temperature at every point, and change over time on the project map — two dates of one site compared, photo by photo and layer by layer.
The model library on this page — number plates, fire and smoke, crowds, vessels — and one auditable record per finding are the direction of the build, not yet in the product.
Talk about your site
Every facility has a different definition of "something worth knowing about". Tell us yours.