A camera with enough pixels on the plate
What matters is not the camera's total megapixels but how many pixels land on the plate. Aim for roughly 130–150 pixels across the plate width.
Home / The hardware behind license plate recognition
LPR · HardwareThe hardware behind license plate recognition.
You do not need a specialist system to read plates. You need the right camera settings, a good mount and a PC with enough power.
Free LPR software · one camera free for life · plate recognition included · about 30 MB
Works withWindows 10 & 11ONVIF camerasHikvisionDahuaReolinkAxisNvidia GPU
LPR · Checklist
What a reliable LPR setup needs.
Even the best algorithm cannot read a plate it cannot see. These are the parts that decide it.
A fast shutter
Moving cars blur. A shutter of 1/500 s suits slow gates and car parks; 1/1000 s or faster is needed for traffic at speed.
Infrared for night
Infrared cameras see in total darkness (0 lux). Plates reflect IR strongly, so they stand out even when headlights would otherwise wash them out.
An IR illuminator for distance
Built-in IR LEDs typically reach 15–30 ft (5–10 m). A separate IR illuminator extends that for longer driveways and wide entrances.
The right lens
A varifocal or telephoto lens lets you zoom in on the lane, so the plate is large in the frame instead of a speck in a wide view.
A PC to run the software
Windows 10 or 11, 64-bit, 8 GB of memory. A graphics card with 4 GB — ideally Nvidia — speeds up AI recognition; without one the processor does the work.
LPR · Night
Headlight glare and night reads.
At night a car's headlights can blind a normal camera, and the plate beneath them turns into a white blur. Two things fix it:
- Infrared illumination with an IR-sensitive camera. The human eye sees 400–700 nm; IR cameras see beyond that, and plate material reflects it back brightly.
- A short exposure so the bright headlights do not overwhelm the frame, which also removes motion blur.
TipAim the camera at the rear plate where you can. Rear plates are not behind headlights, so night reads are far easier.
LPR · Then and now
From overnight batches to real time.
Early systems were limited by processor speed and memory. They recorded video first and analysed it later — sometimes 48 hours later — because the computers of the time could not keep up.
Today, plates are read the moment the vehicle passes. Real-time reading is what makes the modern uses possible: opening a gate for a known car, alerting on a vehicle of interest, or logging every visitor as they arrive.
What separates a good LPR system
- High read rate in bad conditions: night, rain, angles and dirty plates.
- Support for many countries' plate formats.
- A searchable log with a picture of each vehicle.
- Alerts and actions — email, gate, lights — on known or unknown plates.
- Works with the cameras you already own, over the open ONVIF standard.
LPR · Questions
Hardware: quick answers.
What resolution camera do I need for license plate recognition?
A 2–4 MP camera is plenty for one lane if it is zoomed so the plate is about 130–150 pixels wide. Higher resolution helps when one camera covers several lanes.
What shutter speed is best for LPR?
About 1/500 s for slow-moving cars at gates and car parks, and 1/1000 s or faster for road traffic.
Do I need a graphics card?
No, but it helps. An Nvidia card with 4 GB lets AI recognition run faster and handle more cameras; without one, the CPU does the work.