Automatic number plate recognition is a system, not a single camera. Below is the Hikvision AcuSense NVR range we hold — the DS-7608NXI-K1/8P and DS-7616NXI-K2/16P(E) are the two whose own specification names ANPR under camera-side analytics — and the guidance covers what has to be true at the lane before any of it reads a plate.
An ANPR install is a chain, and every link has to hold: a camera aimed down the lane, an illuminator that suits the plate rather than the scene, a recorder that can receive and search the events, storage sized for the retention you promised, and — where the plate opens something — a barrier or access controller taking the relay. The part most specifications get wrong is where the intelligence sits. It is in the camera, not the box in the cabinet. Our own catalogue text for the Hikvision DS-7616NXI-K2/16P(E) is explicit: "Camera-side analytics, including ANPR and AcuSearch, require compatible cameras with the applicable functions enabled." A recorder on its own will never read a plate.
That is why the recorder still matters. Where our own Hikvision spec text names ANPR, it is on the NXI K-series: the DS-7608NXI-K1/8P carries a spec row headed "AI by Camera" whose value reads "perimeter protection, throwing objects from building, motion detection2.0, ANPR, VCA", and the DS-7616NXI-K2/16P(E) names it in the sentence quoted above. What that buys you is a recorder that accepts the plate events the camera generates, holds them against the footage and lets you search on them later. Pick the channel count for the system you will grow into, then check the PoE budget against the cameras you intend to hang off it, because an ANPR camera carrying its own illuminator is not a light load.
Then there is geometry, which decides whether any of it works. Our lens calculator states the DORI thresholds from IEC 62676-4: 250 pixels per metre to identify a stranger, 125 to recognise someone you already know, 62 to observe what happened, 25 to detect movement. Those are PERSON thresholds, and a plate is a harder target than a face: OCR wants something like 100 to 130 pixels across the 520 mm plate, which is 200 to 250 px/m AT THE PLATE — the Identify end of the scale, not Recognise. Work the lens, the standoff and the lane width out against that figure before the ladder comes out. Angle matters as much: aim down a lane rather than across it, because the further the plate is skewed from square-on, the fewer characters the OCR gets. Night is the other half of the problem. UK number plates are retroreflective — they throw light straight back at its source — so a general-purpose IR camera lighting the whole scene blows the plate to white and reads nothing, which is why a genuine ANPR camera pairs narrow-beam IR with a shutter fast enough to freeze a moving vehicle.
Two things to settle before the order. First, compliance: reading number plates is processing personal data under UK GDPR, so the ICO expects a data protection impact assessment before deployment, the minimum number of cameras for the stated purpose, and signage that says cameras are in use and names who operates them. Their ANPR guidance at ico.org.uk is the document to read, and on a client site it is worth agreeing in writing who the data controller is. Second, sourcing: we are a trade distributor, not an installer, and an ANPR camera is a special-order line rather than a shelf item. The recorders, PoE switches, surveillance storage and cable an ANPR install needs are held here; the camera we source to the lane. Send us the standoff, the lane width and the mounting position and we will specify against them.
You need a recorder that can receive, store and search camera-side ANPR events. Our catalogue text for the Hikvision DS-7608NXI-K1/8P lists ANPR under "AI by Camera", which is the distinction that matters: the analytics run on the camera, and the recorder holds and indexes the results. A recorder on its own will not read a plate, whatever it is connected to.
In daylight, at a short standoff, square-on, sometimes. At night, reliably, no. UK plates are retroreflective and a general-purpose IR illuminator overexposes them to a white rectangle. Plate reading needs roughly 200 to 250 pixels per metre at the plate — about 100 to 130 pixels across a 520 mm plate, which is the Identify end of the DORI scale rather than Recognise — plus a shutter fast enough to freeze the vehicle. A general-purpose camera is not set up to do both at once.
Shallow, horizontally and vertically. Every degree the plate is skewed away from square-on to the lens costs the OCR usable characters. Aim down a lane rather than across it, and keep the vertical angle as low as the mounting position allows — which usually means mounting lower and further back than instinct suggests.
Reading number plates is processing personal data under UK GDPR. The ICO expects a data protection impact assessment before deployment, the minimum number of cameras needed for the stated purpose, and signage saying cameras are in use and who operates them. Their published ANPR guidance is the reference; on a client's site, agree in writing who the controller is.
Not as a shelf line. We hold the recorders, PoE switches, surveillance storage and cable an ANPR install needs, and source the camera to the site requirement. Open a trade account and send us the lane details — standoff, lane width, mounting position and whether it is entry-only or entry and exit.
5 products
