It is 8:55 in the morning and raining. The barrier does not open for the car in front, and a queue builds up behind it. The guard lifts the barrier by hand, and this happens ten times a day — until everyone is convinced that the plate recognition does not work. In fact, the technology works. The installation does not: the camera is mounted too high, the angle is too steep, and the car drives through the read zone before it stops.
Automatic number plate recognition (ANPR, also called LPR) is one of the most convenient tools for parking and site-entry control — when it is used in the right place with the right expectations. This article explains what its accuracy depends on, where it is strong, where it is weak, and how to make it part of an access control system.
How ANPR works
The system goes through several steps: the camera captures a frame, software locates the plate in it, recognizes the characters, compares the result with a list and, if the plate is permitted, sends a command to the barrier or gate controller. Every event — with image, time and result — is logged. Both dedicated ANPR cameras with built-in infrared illumination and analytics and ordinary cameras with recognition software running on a server are used. Dedicated cameras usually give more stable results in difficult lighting.
It also matters which mode the system will run in. The simplest is a whitelist: the barrier opens only for registered plates, and an operator decides on everything else. In paid parking, the system matches entry and exit plates and calculates the length of stay. Sometimes ANPR does not control a barrier at all and only keeps a log — for example, to record truck arrivals. The mode directly determines how critical an individual error is: a missed log entry is a nuisance, while a car stuck at the barrier is an immediate problem.
Geometry is decisive
Recognition accuracy is determined above all by how the camera sees the plate. In practice, a few rules apply:
- Angle: the camera should look at the plate as squarely as possible. The larger the horizontal or vertical angle, the more distorted the characters. Manufacturers state a maximum permissible angle — respect it.
- Plate size in the frame: a minimum number of pixels across the plate width is required. It must be met at the actual read point, not on average across the frame.
- Read zone: the best results come where the vehicle moves slowly or stops — in front of a barrier, after a bend, at a speed bump.
- Speed and exposure: a fast-moving vehicle needs a short exposure, which in turn requires good illumination.
- Lighting: infrared illumination brings out reflective plates well at night and reduces headlight glare; by day, low sun and shadows are the main challenge.
One important detail: on entry, the camera usually reads the front plate; on exit, the rear one. If some vehicles have no front plate or a damaged one, the entrance will show it immediately.
Where ANPR works well
ANPR works best at controlled entrances where vehicles stop at a barrier and traffic is predictable: staff parking at an office, a residential parking lot, a hotel where guests pre-register their plate, or a logistics yard where arrival times and unloading bays are recorded automatically by plate. In such cases the system removes the hassle of issuing and losing cards and creates an accurate log of who came and went, and when.
Where ANPR fails
No ANPR system reads every plate. It is important to know in advance which situations are problematic:
- Dirty, damaged, faded or non-standard plates — including decorative frames that cover characters.
- Foreign plates with different fonts and formats, if the software is not configured to recognize them.
- Motorcycles, which only have a rear plate, and trailers whose plate does not match the towing vehicle.
- Plates obscured by bike racks, tow bars or cargo.
- Two vehicles following closely — the second may pass while the barrier is still open.
- Weather and lighting: heavy rain, snow, fog, low sun in the lens, direct headlight glare.
That is why every ANPR entrance needs a plan for the no-read case: an intercom to the guard post, a backup card reader, or remote opening by an operator — not a queue of cars.
A plate identifies the vehicle, not the driver
ANPR recognizes a number plate — not a vehicle, and certainly not the person inside it. Plates can be cloned, and anyone can arrive in a permitted car. For convenience this is usually acceptable. For high-security areas, however, ANPR should be one authentication factor among several — together with a card, a PIN or visual confirmation by an operator.
Integration: ANPR as part of a system
ANPR rarely works in isolation. Most often it is connected to the access control system, where an employee's plate is linked to their profile: when an employee leaves and their card is revoked, the plate is revoked automatically too. Other integrations include parking management and payment, guest pre-registration, an intercom for exceptions, and the video surveillance system, where each event is linked to the corresponding recording.
Technically, this means a network connection to the gate (often fiber, if the gate is far from the building), stable power, a relay or network interface to the barrier controller, and one shared database where permitted plates are managed once.
Pilot testing and acceptance criteria
- Install the camera at the real location and test it by day, at night, in rain and with low sun.
- Measure the read rate on real traffic: of all vehicles, how many were recognized correctly on the first attempt.
- Record error types separately: no read at all, or a wrong read — the second is more dangerous.
- Check what happens in the no-read case and how long the operator takes to decide.
- Make sure adding and removing plates is simple, and that it is clearly defined who does it.
Agree the target read rate and how it will be measured with the installer in advance — over a representative period, including peak hours and night. The accuracy quoted in a manufacturer brochure was achieved in ideal conditions; the only figure that matters to you is the one measured at your entrance, on your vehicles.
If you are planning to automate parking or site entry, get in touch — we will assess the geometry and conditions on site and tell you how well ANPR suits your entrance.