Roads & Highways · AI road condition survey for concessionaires and contractors

Every patrol becomes located, actionable evidence.

RailEyes fits the vehicles already driving your network, such as patrol vans and maintenance crews. Each pass becomes a record of which kilometres were observed, when, by which vehicle, and what was visible.

On a real drive · A-4 Autovía del Sur, 19 October 2026

Pothole and crack detection, from a patrol pass to the map.

One pass on the A-4: anonymised at capture, calibrated, road damage detected, a jolt picked up from the camera's own motion, and the deteriorated stretch placed on the map.

A-4 Autovía del Sur, RailEyes drive · 62 s, no sound · anonymised at capture

The problem

Your contract measures the road. Your evidence is scattered.

Condition and response times are contractual.

Many concession and maintenance contracts set condition indicators, intervention deadlines and penalties. Meeting them is half the work; showing that you met them is the other half.

Patrols run every day, but leave little structured evidence.

The network is driven constantly, yet what was seen often ends up in notes, phone photos and memory.

Survey campaigns are periodic; visible condition is not.

Between campaigns, signs get hidden by growth, barriers get struck, debris appears and potholes open.

Claims and disputes need dates.

Knowing what a stretch looked like at a given time and place settles questions that otherwise take days to answer.

What you get

What RailEyes delivers on your network.

Coverage and recency

Which kilometres of each carriageway were observed, when, and by which vehicle, and which stretches have not been observed for too long.

Route matching

Every observation placed on the right road and carriageway, merged into one pass per stretch per vehicle.

Potholes and cracking

Visible surface defects, each with image evidence and a location.

Signs

Sign detection, and signs whose visibility appears obstructed.

Ride jolts

Jolts that shake the vehicle, picked up from the camera's own motion and placed on the chainage, where a pothole or a failed patch often sits.

Corridor imagery and diagrams

Top-down imagery of the corridor at 2 cm per pixel, with each defect measured and every 20 m graded on a straight-line diagram.

Calibrated measurements

Approximate size and distance from a single camera after a per-installation depth calibration: how large a defect is, how close growth is to a sign.

Change between passes

New, worsening and repaired findings, compared pass by pass on the same stretch.

Roadside vegetation

Growth near signs, sightlines and barriers, building on the vegetation model used on rail.

Anonymisation

Faces and number plates blurred automatically before review, so footage of public roads can be used under GDPR.

Your chainage and exports

Observations referenced to your own kilometre points and asset inventory, exported as CSV or GeoJSON.

What RailEyes does not do

It does not replace laser pavement surveys: roughness (IRI), rutting and texture stay with specialist survey vehicles. RailEyes covers the visible condition between campaigns and shows where to send them. It is not a safety system and does not make automated decisions.

From one pass · A-4, October 2026

A corridor survey from a single patrol pass.

The same pass, rebuilt in 3D and laid flat: top-down imagery of the carriageway, aligned on its own lane markings, every defect measured, and each 20 m graded on a straight-line diagram together with the ride shake.

Top-down carriageway imagery · defects tinted by type
2 cm
per pixel, top-down imagery of the carriageway
12
defects measured: type, area, lane and chainage
20 m
graded sections · worst: 14 % of the carriageway damaged
≈ 4×
the normal shake at the jolt, placed on the chainage
Straight-line diagram of the surveyed corridor: imagery, defects by type, condition grade per 20 m and ride shake along the chainage.
Straight-line diagram: imagery, defects, condition per 20 m and ride shake along the corridor

How it works

From a patrol pass to a reviewed finding.

  1. Step 1

    Capture

    A RailEyes camera on a vehicle that already drives the network. The road version runs on a standard 12 V vehicle supply. Footage from compatible dashcams can also be imported in bulk.

  2. Step 2

    Upload

    Over mobile data, through a small gateway in the vehicle, or through a hub at your maintenance centre that uploads when vehicles return. Chosen per network.

  3. Step 3

    Anonymise

    Faces and number plates are blurred before anyone reviews the footage.

  4. Step 4

    Locate

    GPS and motion data are matched to the road network, by carriageway and by pass.

  5. Step 5

    Detect and review

    Findings appear on a map with their image evidence. Your team confirms or rejects them, and those decisions improve the models.

Why RailEyes

Built in service, not in a lab.

  • Proven on scheduled trains. RailEyes units have ridden more than 1 million km of scheduled service, with about 6 TB processed on board, through a completed paid contract with Comboios de Portugal and Infraestruturas de Portugal.
  • One platform for rail and road. For operators with both kinds of asset, the same capture, location matching and review workflow covers both networks.
  • Image quality under our control. We design the camera, optics and firmware, so evidence does not depend on whatever camera a vehicle happens to carry.
  • Measurements, not just flags. A per-installation depth calibration turns a single camera into an approximate measuring tool.
  • Privacy by design, for European rules. Anonymisation is part of the pipeline, with flexible deployment: on-premise, cloud or hybrid.

See our references · EU AI Act and privacy

Start with one corridor

A pilot you can measure.

  1. 1. Define

    Agree the corridor, host vehicles, target conditions, privacy requirements and success criteria.

  2. 2. Equip and capture

    Install on your patrol or maintenance vehicles. Your operations stay unchanged.

  3. 3. Review

    Your team works with mapped findings and coverage for the pilot period.

  4. 4. Decide

    Measure the results against the agreed criteria and against your current method, then decide whether to expand.

Success criteria we propose measuring

  • Kilometres observed per week, per carriageway
  • Maximum time since last observation
  • Share of findings accepted by your reviewers
  • Reviewer time per finding
  • Cost per kilometre observed, compared with your current method

Questions operators ask

Road condition surveys: frequently asked.

Do we need new vehicles?

No. RailEyes fits the vehicles that already drive your network.

Does it replace our pavement surveys?

No. Laser surveys measure what a camera cannot, such as roughness, rutting and texture. RailEyes covers visible condition between campaigns and helps target them.

What about privacy?

Faces and number plates are blurred before anyone reviews the footage. Hosting and retention are agreed with you.

Do we need IT integration to start?

No. A pilot runs on RailEyes infrastructure. Integration with your asset or maintenance systems can follow if you want it.

Who owns the data?

Data ownership, hosting location and retention are set out in your contract before any capture starts.

What does a corridor survey include?

A top-down image of the carriageway aligned on its lane markings, every defect with its type, area, lane and chainage, condition graded in 20 m sections, a straight-line diagram and the ride jolts along the route.

We manage city streets, not motorways. Is there an offer for us?

Yes. Cities and regional road authorities can join the RailEyes Smart Cities Programme: a managed street survey, a report and a workshop. See the Smart Cities Programme.