How many Flock ALPR cameras has SIREN mapped?

Where are the mapped Flock ALPR cameras? 115,306

Every camera is binned into a cell 0.25° on a side — about 28 km of latitude — and each occupied cell is drawn as its own projected quadrilateral, shaded by how many are in it. Land with no mapped camera is hatched, not left blank, because a blank would be an assertion this data cannot make. The picture needs no JavaScript and every figure in it is a number in the tables below. With it, the map pans and zooms, reads out any cell you point at, and — once you are in close — offers the individual cameras, each with the way its lens points.

Which jurisdictions have mapped Flock cameras? 52/56

Ranked by cameras mapped. Read the counties with none mapped column beside every row: California tops the ranking with 16,891 and still has 5 counties nobody has mapped. A state that looks saturated and a state that looks empty may differ only in how many people were looking.

All 56 states and territories in the roster, ranked by cameras mapped, as mapped in OpenStreetMap on 2026-09-26. The zero rows are printed, not omitted: a state with none mapped is a state nobody has mapped. The fourth column is the one to read sceptically — it is the count of counties in that state where the answer is unknown.

Which counties have the most mapped cameras?

The 25 counties with the most cameras mapped in them, of 3,235 in the roster, as mapped in OpenStreetMap on 2026-09-26. 948 counties have none mapped and are not a tail of this ranking — they are outside it, because nobody has looked.

Which way do the cameras look? 113,214

The bearing the camera looks along, as tagged. It is the field nobody renders and the one that says most about what a camera is for: an ALPR aimed along a carriageway reads the traffic on it. Cameras with no direction tag are absent from these bins entirely — they are not binned as north. Units: degrees clockwise from true north; 0 is due north.

* The north bin is inflated and nothing here hides it. A bare direction=0 means due north AND is the value left behind when a bearing was never established. The two cannot be told apart, so the north bin is inflated by an unknown amount. Nothing is reweighted to hide it. 5,741 cameras — 5.1% of those with a parsed bearing — are in that state.

The untagged are not in the rose. 2,394 cameras have no usable bearing (2,367 carry no direction tag at all and 27 carry one this collector could not parse). They are absent from every wedge above rather than binned as north, which is why the rose totals 113,214 and not 115,608.

How was the bearing written?

OpenStreetMap accepts 4 forms in this field and they are not equally solid. One of them is derived rather than tagged, and a page that folded them all together would be publishing a computation as an observation.

• 105,107 bearing a plain bearing in degrees, e.g. "142".
• 4,141 multi an OSM multi-value, e.g. "0;0". The first value is used; the others are not averaged in.
• 3,908 sector an OSM field-of-view arc, e.g. "338-23" — 45 degrees wide, centred on 0. The centre is DERIVED, not tagged.
• 58 compass a 16-point compass letter, e.g. "NNE".

4,053 cameras state a field of view as well as a bearing — most commonly 45° wide, on 3,650 of them. Only the sector form states a field of view. Where it does, the width is real; where it does not, no width is inferred. Widths are not carried per camera in flock-points.json — they are aggregate only.

How complete is the Flock tag data?What the tags carry

A tag is only present if a volunteer typed it. This is the completeness of each field across the 115,608 mapped objects — the honest denominator for every figure on this page, and the reason the operator ranking below is a fact about attribution rather than about deployment.

• direction 98.0%

  113,241 of 115,608 carry it; 2,367 do not. Which way the lens points. The field this page is built on.

• camera:type 97.9%

  113,199 of 115,608 carry it; 2,409 do not. Fixed, dome, panning. Almost always fixed.

• surveillance 88.6%

  102,466 of 115,608 carry it; 13,142 do not. Whether the camera watches a public or private space.

• surveillance:zone 87.9%

  101,680 of 115,608 carry it; 13,928 do not. What it is pointed at. Almost always traffic.

• camera:mount 28.3%

  32,759 of 115,608 carry it; 82,849 do not. Pole, street lamp, traffic signal, wall.

• operator / operator:type 12.9%

  14,953 of 115,608 carry it; 100,655 do not. Who runs it. Recorded for one camera in eight.

Who is recorded as operating them? 12.9%

Do not read this as a deployment ranking. Only 14,953 of 115,608 cameras record an operator at all. A chain store's cameras are easy for a passing mapper to attribute; a municipality's usually are not. This table measures who is easy to attribute, which is a different thing from who deploys the most.

The 12 operators named most often, of the 14,953 cameras that name one at all — 12.9% of those mapped. Shares below are of that 14,953, never of the whole.

What is this Flock register not?

Published by the collector, printed here unedited.

• Every figure here counts cameras as mapped in OpenStreetMap on 2026-09-26. It is not a count of cameras that exist.
• 948 of 3,235 counties have zero mapped cameras. That means nobody has mapped them, not that they have none.
• OpenStreetMap coverage is volunteer-driven and uneven; the map is partly a map of where mappers have been.
• Direction is tagged on 113,241 of 115,608 cameras. The untagged ones are shown as untagged, never as facing north.
• 5,741 cameras carry a bare direction=0. That is both "due north" and what a mapper leaves behind when the bearing is unknown, so the north bin is inflated by an amount nobody can measure. It is not reweighted.
• An operating agency is recorded for 14,953 of 115,608. For the rest OSM does not say who runs the camera, and neither does this page.
• County assignment uses TIGER polygons simplified to about 550 m, so a camera within a few hundred metres of a county line can land in the neighbouring county.
• 258 mapped cameras fall outside any US county. They are counted and kept, not dropped.
• Flock Safety publishes different deployment figures. The two are not reconciled here, because reconciling them would mean imputing.

This is a snapshot and not a time series. Nothing here says when a camera was installed or whether it is switched on — only that an OpenStreetMap object exists. A camera taken down keeps its tag until somebody visits and removes it.

Where did the Flock map data come from?

OpenStreetMap ALPR nodes tagged Flock Safety, via Overpass. Every camera here is an OpenStreetMap object tagged man_made=surveillance and surveillance:type=ALPR, carrying at least one of manufacturer=Flock Safety, brand=Flock Safety, operator=Flock Safety. The three are unioned rather than taken one at a time, because they overlap: 1,652 objects carry more than one of them and would otherwise be counted twice. OpenStreetMap's own snapshot clock for this extract reads 2026-09-26T03:52:04Z; this page was generated 2026-09-26 03:58:53 UTC.

Substitute a bounding box into query_template and POST it to the endpoint as a form-encoded `data=` field (a raw POST body answers HTTP 406). Every figure on this page is the arithmetic in collector/flock.mjs over that response and nothing else.

How was the harvest tiled?

A single United States query is about 42 MB and times out, so the planet was tiled: 5 seed boxes covering it exactly once with no gap and no overlap, the contiguous states pre-cut into a grid. 36 data queries and 5 count queries were issued to one endpoint; 0 tiles needed splitting, 0 levels of recursion were used and 0 duplicates were absorbed at tile edges.

Tiles are fetched with no `out` limit, so a tile either arrives whole or fails loudly — there is no silent truncation to detect. One count query per seed box cross-checks the result: seed_cross_checks compares the box total against the distinct cameras its tiles yielded, and a shortfall over 2% aborts the run rather than publishing a partial map.

The integrity check, one row per seed box: what the server counted before the fetch against what the fetch actually yielded.

The conus row collects two more than it counted. That is not an error and it is not rounded away: OpenStreetMap was edited between the count query and the fetch, and the two cameras added in that window are real.

How was geography assigned?

Counties and states come from Census TIGERweb county boundaries, 3,235 county polygons simplified to about 0.005° — roughly 550 m. 3,314 cameras sit close enough to a county line that the assignment could belong to the neighbour, and 258 fall outside any US county and are kept rather than dropped.

What is the licence?

© OpenStreetMap contributors — https://www.openstreetmap.org/copyright. Open Database License (ODbL) v1.0. ODbL requires this attribution and the link wherever the data is shown. It is a licence condition, not a courtesy.

How do you take the data?

Both files are published beside this page, byte-identical to what the collector wrote — the build copies them verbatim rather than re-serialising, so the checksum of the file in the repository is the checksum of the file served here.

• /api/flock.json Every number on this page: the totals, the coverage block, all 3,235 counties including the 948 with none mapped, and the 5,029-cell grid this map is drawn from.
• /api/flock-points.json 115,608 individual cameras as four packed integer arrays — latitude, longitude, bearing and state. 2,672 KiB. A bearing of −1 means no usable direction tag, not north. The page never inlines it. The map offers to fetch it once you are zoomed past 12× — a button, not an ambush, because 0.9 MB over the wire is the reader's decision — and past 40× draws each camera with its bearing.

How do you recompute it?

Substitute a bounding box for the four zeroes and POST it to https://overpass-api.de/api/interpreter as a form-encoded data= field. A raw POST body answers HTTP 406.

[out:json][timeout:600];
(
  nwr["man_made"="surveillance"]["surveillance:type"="ALPR"]["manufacturer"="Flock Safety"](0.00000,0.00000,0.00000,0.00000);
  nwr["man_made"="surveillance"]["surveillance:type"="ALPR"]["brand"="Flock Safety"](0.00000,0.00000,0.00000,0.00000);
  nwr["man_made"="surveillance"]["surveillance:type"="ALPR"]["operator"="Flock Safety"](0.00000,0.00000,0.00000,0.00000);
);
out tags center qt;

Source state live · refreshed every 7 days · no API key · page generated 2026-09-26 03:58:53 UTC · built by collector/flock.mjs.