What it is, in plain language
Parkomate's ANPR runs natively on the same Android POS handhelds that the operator already carries, using Google's on-device ML Kit. Plates are read in milliseconds without a network round-trip — which means it keeps working when the internet drops and there's no per-scan cloud cost. Three capture modes: auto (live frame), manual (capture button), and keypad (operator types the plate as fallback).
Key features
Specification highlights drawn directly from engineering documentation. Full BOQ provided after on-site survey.
Technical architecture
The ML model is bundled into the operator app and updated through the same channel as the rest of the app — no separate model server. Reads happen on the device CPU/NPU in under 200 ms typical. The operator can choose continuous auto-capture, single-shot manual capture, or fall back to keypad entry if the plate is damaged or obscured. Every read writes a timestamped, hashed audit row that survives the immutable audit trail.
Why customers choose this system
The measurable operational and customer-experience improvements you can expect after deployment.