Peregrine Data
Street-level road data from fleets in daily service
Cameras read signs, surface damage and roadworks as vehicles drive their normal routes. You get georeferenced records, ready for your GIS.
- No survey campaignThe driving is happening anyway
- Quarterly refreshFaster where fleet density allows
- JSON, CSV, GISReady for ArcGIS on arrival
The map is only the bottom layer
Four layers of road data, from permanent geometry to live traffic
You already have the base map. Peregrine supplies the three above it.
Highly dynamic data
(vehicles, pedestrians)
Transient dynamic data
(congestion, signal phase)
Transient static data
(roadside infrastructure)
Permanent static data
(map data)
Every record carries a type, a subtype, a position, a heading and the date it was seen.
Who uses the data
Three groups build products on top of it. One uses it to run a network.
Today
Survey vehicles pass once or twice a year, and minor road classes are covered late or not at all.
With Peregrine
Changes surface between campaigns, at a far lower cost per kilometre.
Today
Devices cover enormous distances daily and produce nothing beyond the safety product they were sold for.
With Peregrine
A geodata layer you can sell on, from hardware your customers already run.
Today
Operational design domains are validated on a handful of instrumented routes.
With Peregrine
Objects, manoeuvres and seasonal change across real routes, as GIS layers or ASAM OpenSCENARIO.
Today
Manual inspection is slow and partial, so maintenance is planned on an incomplete picture.
With Peregrine
Fleets on their normal rounds surface damage and missing signage while repairs are still cheap.
USE CASES
See what’s happening on your roads
Street-level intelligence from cameras, ready for your use case.
Traffic signs
Detect, classify and localise signs so your map stays level with the road.
Construction sites
Active works, barriers and lane changes, picked up as they appear and as they clear.
Road surface damages
Potholes, cracks and surface wear, scored for severity so repairs can be ranked.
Waste bins
Locate and check bins across the network to plan collection and maintenance rounds.
Bus lane violations
Vehicles driving or parked where they should not be, and when it happens.
ODD monitoring
Real-world driving conditions across seasons and weather, for operational design domains.
Coverage of specific classes varies by region and fleet density. Tell us the use case and we will confirm what is reachable on your network.
Mobile mapping, without the mapping fleet
The question is not how often someone drives past. It is how out of date your map is on the day you need it.
Illustrative. Actual pass frequency depends on fleet density in your area.
Coverage
Delivery vans and municipal vehicles reach the residential streets survey routes skip.
Frequency
The same segments are seen again and again, so a change is dated to a window of days rather than a year.
Cost per kilometre
No survey vehicle, no driver, no campaign to plan. The cost sits in the camera that is already in the windscreen.
A survey programme is a fixed cost you repeat. Fleet sourcing is a marginal cost on driving that is already paid for, which is what moves the cost per kilometre and the refresh rate at the same time.
From a frame to a point on your map
A detection is not yet geodata. It becomes geodata once it has a position on the ground.
Sofia, road asset sourcing. A fleet vehicle on an ordinary route. Every marker you see was produced on the device as it drove.
See the object
Detected and classified in the frame, on the camera's own processor.
2Track the camera
Visual localisation follows the vehicle's own path, which holds up where GPS is weak.
52.5200, 13.4050 · 219° 3Place it on the network
Seen from several viewpoints, the object resolves to a position, a heading and a date.
Tell us about your network
Three questions and your details. We will come back with the coverage we can reach on your roads and what a first delivery looks like.
- 1. We check what fleet coverage exists on your roads today.
- 2. You get a written answer with the cadence it supports.
- 3. If it works, a sample of your own area before anything is signed.
See it on a real network
Road Insights over Berlin Mitte. Every marker was produced by a fleet vehicle on an ordinary route.
- Traffic signs
- Traffic lights
- Road damage
- Route driven
Built to integrate
One record per observation, in the format your team already works in. No proprietary viewer, no bespoke pipeline, nothing to reverse engineer.
{
"type": "traffic_sign",
"subtype": "speed_limit_30",
"latitude": 52.5200,
"longitude": 13.4050,
"heading": 219.105,
"last_observation": "2026-09-02"
}GIS layers
Straight into ArcGIS or any common GIS platform.
Flat files
Scheduled .csv or .json drops on your refresh cadence.
REST API
Query by area, class or date. Pull only what changed.
ASAM OpenSCENARIO
For AV and simulation teams working with manoeuvres.
What your engineers get on day one
- A stable schemaEvery record carries the same fields, whatever the asset class. Add a layer and nothing downstream breaks.
- Query by area, class and dateAsk for a bounding box, a set of asset types, or only what has changed since your last call.
- Delta-firstPull the change set rather than the whole network, so a refresh is cheap to ingest.
- A sandbox before you commitWork against a sample of your own area so integration is proven before anything is signed.
In the field
Public sector and research deployments where fleet-sourced observations ran a real network.
16,000 km monitoredPotholes found before they spread
Municipal trucks and delivery vans on their normal routes surfaced road wear across the city network.
Read the case
20 vehiclesParcel rounds that map the city
Hermes fitted 20 delivery vehicles with Peregrine cameras over a winter, then evaluated the data with map providers and cities.
Read the interview
The scene, before arrival
Detection overlays streamed into fire brigade vehicles en route.
Read the caseSelected by leading deep tech programs globally, and collaborating with leading universities in Germany in robotics and edge AI.
Fully GDPR compliant
- Records, not recordingsA sign becomes a row with a type, a position and a date.
- Faces and plates blurredAnonymised automatically before scenes are stored.
- EU-based serversStorage and processing stay in the EU, imagery encrypted at rest.
Data layers you can order
Start with road surface condition and add the layers your use case needs.
Road surface damages
Four damage types, localised and scored for severity, so a maintenance list arrives ranked rather than raw.
- Pothole
- Transversal crack
- Longitudinal crack
- Alligator crack
- RefreshQuarterly
- Road network100 km per unit
- Service period12 months
- Devices included5
- Delivered asJSON or CSV
Add the layers you need
Coverage and cadence depend on fleet density in your area. Send us your network and we will come back with what is achievable.