ANPR

SmartCity ANPR Engine

The recognition engine: detection, plate localisation and OCR.

Licensing: perpetual
Latest version: 5.2.0
SKU: CI-ANPR-ENG
Delivery: Certified local integrator

Pricing

Public price‏2,400.00 US$/licence

Partner pricing is visible after sign-in. Project pricing is set by your local integrator.

The ANPR Engine is the component that actually reads plates. It takes a video stream or a triggered image, locates the vehicle and the plate, and produces a plate string with a confidence value and the region format it matched.

Regional plate formats

The engine ships with plate-format profiles for the regions it supports, including GCC formats. Format profiles matter: a plate layout the engine has no profile for will read poorly regardless of image quality.

Sizing

Throughput depends on the hardware, the number of streams and whether GPU acceleration is available. Your integrator sizes the engine against the lane count and the expected vehicle flow.

Capabilities

Vehicle and plate detection

Locates the vehicle and the plate region before OCR runs, which is what makes free-flow lanes viable.

Regional plate profiles

Format profiles per country so the engine knows what a valid plate looks like in that jurisdiction.

GPU acceleration

Optional GPU inference for higher stream density on the same server.

Confidence scoring

Every read carries a confidence value so downstream rules can require a threshold.

Triggered and free-flow modes

Works with loop or beam triggers, or continuously against a live stream.

Specifications

Specifications are maintained from the product database. Deployment-specific figures depend on hardware, camera placement and network conditions — your integrator will size the system.

Runtime

Operating system	Windows Server 2019+ / Ubuntu LTS
Acceleration	CPU or CUDA-capable GPU

Input

Sources	RTSP stream, triggered image, folder watch

Output

Event fields	Plate, confidence, region format, timestamp, lane, images

Licensing

Licensed by	Stream

Verified integrations

RTSP streams
LPR cameras
Loop and beam triggers

Release history

5.2.0Latest29 يوليو 2026
Additional GCC plate-format profiles; improved night-time plate localisation.
5.0.0LTS02 أكتوبر 2025
New detection pipeline with GPU inference support. Long-term support release.

Frequently asked questions

What affects read quality most?

Camera geometry, plate illumination and shutter speed — in that order. Software cannot recover a plate the camera never resolved, which is why the site survey matters more than any engine setting.

Does it need a GPU?

No. GPU acceleration increases how many streams one server can handle; small deployments run on CPU.