How do Flock cameras work?

A Flock camera photographs every vehicle that passes it, reads the number plate with optical character recognition, and records the vehicle's make, colour, body type and features such as roof racks or bumper stickers. Each read becomes a timestamped, geotagged record uploaded over cellular to a searchable database.

The sequence, step by step

1. Detection. The camera watches one direction of one road. Motion in the frame triggers a capture.
2. Capture. It takes a still image, usually with infrared illumination so it works at night without a visible flash.
3. Plate reading. Optical character recognition extracts the plate number and the issuing state.
4. Vehicle fingerprinting. This is the part people underestimate. The system also classifies make, body type and colour, and notes distinguishing features — roof racks, bumper stickers, ladder racks, damage. That means a vehicle can be matched even with no plate read at all.
5. Upload. The record — image, plate, vehicle description, timestamp, camera location — goes over a cellular link to cloud storage.
6. Matching. Reads are checked against hot lists such as stolen-vehicle files and warrant lists. A hit generates an alert.
7. Retention. Non-matching reads are kept for a set window, searchable the whole time. See how long the data is kept.

What it is not

Two common misconceptions:

It is not facial recognition. These cameras are aimed at vehicles, not people. Flock Safety has stated its road cameras do not perform facial recognition.

It is not live tracking. There is no continuous trace of your car moving. What exists is a series of discrete sightings: your plate at this pole at this second. The tracking effect emerges from density — enough sightings in enough places and the dots join themselves into a route.

Why the direction of the lens decides everything

A plate reader is a fixed camera pointed one way. It is not a proximity sensor. Three things determine whether it read your plate:

• Bearing. Which way the lens faces.
• Your approach. Whether you drove into that field of view or away from it.
• Distance. Whether you were inside its usable range.

Most plate readers capture the front or rear of vehicles travelling in one direction on one carriageway. Passing a camera on the opposite side of a divided highway frequently produces no read at all.

This is why the OpenStreetMap direction tag matters so much. Of the 127,344 US cameras documented, 121,036 carry a recorded bearing — roughly 95%. Any map that draws a plain dot is throwing away the field that determines whether the camera saw you.

Distance tells you a camera is nearby. Direction tells you whether it was looking at you. Only the second one answers the question you actually have.

Seeing the mechanism in the app

Flock Camera draws the direction as a view cone spreading out from each pin, so the geometry is visible rather than implied:

1. Map tab, zoomed to street level — every camera with a recorded bearing grows a cone showing where it points.
2. Tap a pin — brand, operator, compass bearing (for example "53° NE"), what it watches, and the last verification date.
3. Arm alerts — you get one notification as you come inside your chosen radius, and a second, separate one if you enter the cone itself.
4. History — every camera you passed while alerts were on, each marked "likely captured" or not, based on whether you were inside the cone rather than merely near the pole.

All of that geometry is computed on your phone against a local copy of the database. Your position is never sent anywhere.