Seven parameters that decide whether a plate camera works

1. Plate width in pixels

Recognition needs the characters to be resolvable, which means the plate must occupy enough pixels across. As a working rule, aim for at least 120 to 150 pixels across a standard 520 mm plate for reliable OCR, and more for difficult plates. A 4K sensor spread over an eight-metre-wide lane gives far fewer pixels per plate than the same sensor on a 3.5 m lane. Narrow the field of view before you raise the resolution.

2. Capture distance and lens

Distance and lens are one decision. A longer lens buys plate pixels at the cost of a narrower corridor of capture, so the vehicle must pass through it. Work out the width you need to cover, divide by the pixels you need per plate, and check the lens range on the data sheet can deliver it at your mounting distance.

3. Shutter and motion blur

A plate moving at 30 km/h covers about 8 metres per second. At a 1/500 s shutter it moves roughly 16 mm during the exposure, which is small next to a 520 mm plate; at 1/60 s it moves about 13 cm, which smears characters. Plate capture needs a fast shutter and enough light to support it, which is why the next parameter matters.

4. Illumination and the retro-reflective plate

Plates are retro-reflective by design: they bounce light back toward the source. A dedicated IR or white-light illuminator close to the lens axis uses that property and is the usual reason a night plate is readable when a colour image is not. Check the illuminator range against your capture distance, and check whether white light is acceptable at the site, because it is visible to drivers.

5. Mounting angle

Horizontal angle between the plate and the camera should stay under about 30 degrees, vertical under about 15 degrees. Beyond that, characters foreshorten and OCR accuracy drops even when the image looks sharp to a human. If the lane geometry forces a steep angle, move the camera upstream rather than accepting it.

6. Trigger: always-on or vehicle detection

A camera that records continuously fills storage with empty road. A camera that triggers on a vehicle loop, radar or built-in detection records one image per pass, which keeps the archive small and searchable. For parking and barrier control the trigger also drives the barrier, so its latency matters more than its accuracy.

7. What happens after the capture

Two decisions remain: on-camera recognition or back-end software, and where the list of allowed or blocked plates lives. On-camera recognition keeps the network light and works for a single barrier; a back-end handles multiple lanes, lists and integration with access control or billing. Decide this before choosing the camera, because it changes the interfaces you need on the data sheet.

Manufacturers quote recognition rates from controlled tests. Ask for the conditions: plate standard, speed, angle, light. A 99 percent figure at 20 km/h and a 5-degree angle says little about a 60 km/h gate with a 25-degree offset.