Traffic Enforcement Cameras: Free Speed Camera Map and Data

Traffic enforcement cameras became a legal question in two countries this summer. On 12 July 2026 a decree came into force in Italy that reset which speed cameras may legally issue fines. Only 25 device models are automatically approved under it. Roughly 850 installed devices fell outside that list and cannot currently be used, pending an approval application from each manufacturer. Since 30 November 2025 Italian municipalities have also had to register every device in a national database held by the transport ministry, because a fine from an unregistered device is contestable.

Spain is moving on a parallel track. The reform of the Reglamento General de Circulación takes effect on 1 October 2026 and bans radar detectors outright, at 500 euros and three licence points, with 6,000 euros and six points for jammers, even when the device sits unused in the glovebox.

Two countries, two registries, two sets of rules, and in both cases the same question from drivers and from developers: where are the traffic enforcement cameras, and where does that list come from. There is an answer that predates both decrees and belongs to nobody in particular. Enforcement cameras have been mapped in OpenStreetMap for over fifteen years, and the dataset is larger than most people expect.

We counted it, then counted our own driver-reported database against it so the two could be compared honestly. This article publishes both sets of numbers, the queries that produced them, the vintage of each, and the parts neither dataset can tell you.

What counts as a traffic enforcement camera

A traffic enforcement camera is any camera that enforces a traffic rule automatically, without an officer present. A speed camera is one member of that family, not the whole of it, and the distinction matters as soon as you try to count them, because different countries mean different things by the word.

In the United States the whole category is usually called photo enforcement or automated traffic enforcement. In the UK it is safety cameras. In Germany the everyday word is Blitzer, in Italy autovelox, in France radar, in Spain radar de la DGT. They are the same class of device with different politics attached.

The dataset in this article covers the first four rows, which is where drivers and routing engines need coordinates. The last two are plate readers rather than speed measurement, and they raise a different set of privacy questions that we wrote about in the DeFlock ALPR piece.

One practical distinction worth keeping straight, because it is the most common confusion: a speed camera penalises how fast you were going, a red light camera penalises when you entered the junction, and most modern red light units do both. That is why a single housing can produce either kind of penalty, and why our index keeps them as separate classes rather than merging them into one dot.

What the open map holds today

We counted it directly rather than citing a figure. The numbers below come from the Overpass API on 27 August 2026, against an OpenStreetMap base timestamped 15:06 UTC that day.

Worldwide there are 76,800 nodes tagged highway=speed_camera, plus 24,183 relations of type=enforcement with enforcement=maxspeed, which describe the stretch of road a camera or an average-speed section actually governs.

Method, so you can reproduce or contradict it: each figure is a count of highway=speed_camera nodes inside that country's admin_level=2 boundary. Where the ISO code lookup timed out, we resolved the area by relation id instead. We discarded bounding-box counts entirely, because a box around France returned 11,170 by sweeping in Belgian, Dutch, German, Swiss and Catalan cameras, against 3,928 inside the actual border. We also dropped the United States figure: the query timed out on repeated attempts, and a number we could not verify is worth less than no number.

How that compares with a database nobody can query

An open dataset is only interesting if you know what it is missing, so we ran the same count against our own.

MapMetrics has been collecting enforcement camera reports from drivers for years. The snapshot we counted holds 108,008 unique cameras across 106 countries: 75,258 fixed devices, 27,912 red-light cameras, 3,310 section-control points and 1,528 variable-limit installations. Brazil, Russia, Italy and the United States carry the largest national totals.

The two figures are not the same vintage, and that matters more than the comparison. The OpenStreetMap numbers are live, pulled on 27 August 2026. The driver-reported snapshot was exported in May 2024. Read the table below as a coverage comparison, not as two current measurements of the same afternoon.

Two datasets, one built by mappers editing a public database and one built by drivers reporting from the road:

The interesting result is not that one dataset is bigger. It is that in four of these ten countries the two independent sources land within 10 percent of each other, and in three of them the open map holds more. Two communities that never coordinated, using different tools for different reasons, converged on nearly the same count of German and French cameras. That is about as good a cross-validation as this category of data gets.

Where they diverge, the divergence is worth investigating rather than explaining away. Italy at 2.08 and Portugal at 0.60 are the two ends, and since our snapshot predates the Italian decree by two years, neither ratio can be a reaction to this year's rule changes. The likelier read is simply where each community was active: driver reporting has been dense in Italy for years, and Portuguese mappers have been thorough in a country with fewer reporters. A ratio far from 1.0 is a prompt to go and look, not evidence that either side is wrong.

Method and limits, stated the same way as above. The four collections were deduplicated on coordinates before counting, because the variable-limit collection stores each camera once per speed value, 38,250 rows for 1,528 cameras. Every record carries the same export timestamp, 21 May 2024, rather than a per-camera survey date, so these counts describe what that snapshot held, not when each camera was last seen. A verification field exists in the schema and was never populated. 1,362 points sat outside any country polygon, mostly on coastlines, and are excluded. Ratios against a live OpenStreetMap count therefore understate our side wherever cameras were added after May 2024, and overstate it wherever cameras have since been removed.

How enforcement cameras are tagged in OpenStreetMap

The schema is small, which is exactly why it got populated.

• highway=speed_camera on a node marks the physical device. This is the tag that carries the volume.
• type=enforcement relations carry the meaning. The relation ties a device role member, the camera, to a from and to road segment, so an average-speed corridor of nine kilometres is one relation rather than two disconnected posts.
• enforcement=maxspeed distinguishes speed enforcement from the other values the schema supports, including traffic_signals for red-light cameras, average_speed, toll, and access.
• maxspeed records the limit being enforced, which is what turns a dot on a map into something a routing engine can reason about.
• direction records which way the device faces, so a camera that only catches northbound traffic is not rendered as if it catches both.
• check_date or survey:date, where contributors have added it, tells you when a human last stood in front of the thing. On enforcement data this matters more than on almost any other feature class.

The pattern is the same one that produced the DeFlock ALPR registry: agree a tag, document it, let a community populate it. Nothing about it required permission from a camera vendor or a road authority.

Query it yourself

Paste this into overpass-turbo.eu and run it. It returns every mapped speed camera in Germany, with tags, as geometry you can export to GeoJSON.

[out:json][timeout:180];
area["ISO3166-1"="DE"][admin_level=2]->.country;
(
  node["highway"="speed_camera"](area.country);
  relation["type"="enforcement"]["enforcement"="maxspeed"](area.country);
);
out geom;

Swap DE for FR, IT, ES, NL or any other ISO 3166-1 alpha-2 code. To count rather than fetch, replace the final line with out count;, which is how the table above was produced. For a city-sized extract, replace the area block with a bounding box in (south,west,north,east) order.

Two practical notes. Country-wide queries against the public Overpass instance time out often, so for production work you extract from a Geofabrik regional .osm.pbf and filter locally, or run your own Overpass instance. And the licence is ODbL, which means attribution and share-alike on derived databases, not a free-for-all.

Open data against the official registries

Comparing the two is the most useful thing you can do with this dataset, and it is the comparison nobody publishes.

Spain's DGT states 2,472 fixed radars, 295 section radars, 254 belt and mobile units, and 600 traffic-light cameras. OpenStreetMap holds 2,553 camera nodes in Spain. Those totals are close, but they are not measuring the same population, because the OSM figure mixes fixed, section and traffic-light devices under one tag unless contributors have split them with enforcement relations.

That is the honest reading of every such comparison:

• Official registries record legal status. Whether a device is approved, calibrated, registered, and therefore able to produce an enforceable fine. Italy's ministry database is now precisely this, and after 12 July 2026 an unregistered device is a legal problem rather than a mapping problem.
• Open map data records physical reality. Whether a contributor could see a device from public space and record where it stands. That includes decommissioned housings, empty boxes, and posts whose approval status nobody outside the municipality knows.

Neither is a substitute for the other. A camera that exists physically but is missing from the ministry list is exactly the case Italian drivers are now contesting. A camera on the official list that no longer exists physically is exactly the case a navigation app gets wrong. You need both datasets to see either problem, and only one of them is queryable by anyone.

What the open map cannot tell you

State the limits plainly.

Mobile enforcement is largely absent. A van parked on a bridge for three hours is not a mappable feature, and OSM's verifiability rule correctly excludes it. Depending on the country, mobile units produce a large share of all speeding fines, so a mapped-camera layer systematically understates enforcement. Driver-reported data has the same hole for a different reason: a report of a van that has since driven away describes a moment, not a place.

Freshness varies by region, not by country. Coverage tracks contributor density. A German city with an active local mapping group is close to complete. A rural region three countries away may carry cameras removed in 2019. Read check_date where it exists and treat its absence as unknown rather than as current.

Approval status is not in the data. No OSM tag tells you whether an Italian device sits on the ministry list, or whether a Spanish radar is calibrated. Those facts live in national registries and change without any edit to the map.

Tag discipline is uneven. Some countries map red-light cameras as highway=speed_camera even though a separate value exists. Filter on the enforcement relation when the distinction matters to your use case.

The law is the harder half

Publishing camera locations is legal across the EU. Using a live warning system while driving frequently is not, and the distinction that matters is not map versus app, it is database lookup versus signal detection. Status as of August 2026:

Rules elsewhere differ, sometimes sharply, and they change. Treat the table as the shape of the problem rather than as advice, and confirm the national rule before shipping a driver-facing feature into a market.

For anyone building a product, the practical consequence is that the same dataset supports very different features depending on jurisdiction. A static planning map, a fleet analytics layer, a road-safety study, and a live in-car alert are four different legal objects built from one set of coordinates.

Building it into something

The path from the query above to a working layer is short, and it is mostly the ordinary map problems rather than anything camera-specific.

1. Extract and filter. Pull the regional .osm.pbf, keep highway=speed_camera nodes and type=enforcement relations, and resolve the relation members so an average-speed section becomes a line rather than two points.
2. Snap to the road. A camera node sits beside the carriageway, not on it. Match it to the way it governs, or your distance calculations will be wrong by the width of the road and your direction filter will not work at all. This is a map matching problem.
3. Attach the limit. Carry maxspeed through from the relation, so the layer answers what is enforced rather than only where.
4. Record provenance per feature. Keep the OSM id, the version, and check_date. When a user reports that a camera is gone, you want to know what you knew and when.
5. Render it as a layer, not a dataset. Vector tiles keep a national camera layer at a few megabytes and let you restyle it per market, which matters when the same layer must look like planning information in one country and must not look like a live alert in another.

MapAtlas runs the geocoding, tile and POI infrastructure this sits on, processed in the EU under GDPR, if you would rather build the feature than the pipeline.

Why this keeps happening on the open map

Enforcement cameras are the second instance of the same pattern we wrote about with ALPR mapping, and the mechanism is worth naming because it will happen again.

A category of infrastructure becomes contested. It is visible from public space, so it is verifiable, which is the only test OpenStreetMap really applies. Somebody proposes a small tag, a few contributors populate their own city, and within a few years there is a global registry that no institution commissioned and no vendor controls. Then a regulator changes the rules, as Italy and Spain both did this year, and the open dataset turns out to be the only source anyone can independently check the official one against.

That is the durable value here, and it has little to do with avoiding a fine. Seventy-six thousand cameras were mapped by people who walked past them. The registries that determine which of those cameras may legally issue a fine were compiled by ministries. For the first time, in Italy at least, both lists exist at once, and the interesting work is in the difference between them.