Speeding and tailgating,
caught by the cameras you already have.
AI Speed Radar turns ordinary camera footage into usable traffic evidence. It detects vehicles, estimates speed and following distance, reads the licence plate and builds a sealed case file that cannot be altered unnoticed.
What the system does
Everything below is built and covered by automated tests.
Detection and tracking
Recognises vehicles in the frame and follows them across it. Estimates speed from the calibrated scene and checks following distance against the two-second rule.
Licence plate reading
Built-in plate recognition with regional prefixes and correction of the character mix-ups that always happen. Proven on real photographs and on a moving vehicle filmed at dusk.
Tamper-evident evidence
Every event gets a cryptographic fingerprint of the image and the facts, chained and signed. If anyone changes anything later, it shows. That is what makes the file worth something.
Import and export
A complete case file exports as a ZIP, event lists as CSV. The system can also take in readings from physical radars you already own, so nothing you have bought goes to waste.
Straightforward integration
Ships with its own SDK and more than twenty ready connectors, so it plugs into what you already run instead of demanding that you replace it.
Your data stays yours
Everything is stored locally on your own machine — not in our cloud, not in anybody else's. Retention periods, anonymisation and an access log are built in.
Dashboard and video wall
Sortable event list with filters, user roles, analytics by hour, location and vehicle type, and a wall showing several cameras at once.
The authority stays in charge
Legally sensitive steps — looking up a vehicle owner and issuing a penalty — are performed by the competent authority through an adapter, never by the system on its own.
How we keep the numbers honest
Speed from video is only as good as the camera position and the calibration. So we measure it rather than claim it.
Speed estimation was checked against a public research dataset of 13 cameras and 78 clips with known true speeds. Good camera positions land within 3–7 %; poor ones drift much further, which is exactly why placement matters.
We compared our implementation of the official evaluation method directly against the original research results. The numbers match to the last decimal, so our accuracy figures are not self-graded.
Before a location goes live we run a reference drive and verify it against GPS. If a spot cannot reach the agreed tolerance, we say so and move the camera instead of shipping a number we do not trust.
What it is used for
Traffic analysis
How many vehicles, when the peaks are, where people speed, how often the same plate comes back. The numbers councils actually decide on.
Evidence for enforcement
Detection and a prepared case file that an officer reviews. The decision and the procedure stay with the authority.
Preventive speed displays
Show drivers their speed and collect statistics without penalties — the fastest route to something useful on the street.
Following distance
Distance between vehicles is measured differently from speed and works even where a camera is not calibrated for speed.
How a pilot works
There is no published price here, and the reason is simple: scope differs at every site — how many cameras, how they are mounted, whether on-site calibration is needed. A published number would be invented.
Site visit
We look at the cameras and the road, and tell you straight away whether the location can deliver useful numbers.
Calibration
Reference drive, GPS check, agreed tolerance written down before anything goes live.
Pilot
One location, a fixed scope and a measurable result. If the pilot misses what we promised, it does not continue.
Rollout
More cameras, integration with your systems, training for the people who will use it every day.