ANPR Detection
System

Overview

ASTROC developed a real-time Automatic Number Plate Recognition (ANPR) system capable of detecting, reading, and logging vehicle registration plates from both aerial drone footage and ground-based dashcam video. The system uses custom-trained ASTROC Vision System models for plate detection and ASTROC OCR for character recognition.

Technical Architecture

The ANPR pipeline consists of a two-stage detection system: Stage 1 identifies vehicles in the frame using a general object detection model, while Stage 2 crops the vehicle region and applies a specialised plate detection model for character extraction.

The system runs on edge devices (ASTROC embedded edge compute modules) for real-time inference at 15+ FPS, enabling deployment on moving drone platforms and vehicle-mounted cameras.

Capabilities

The system is highly capable, effectively detecting and reading license plates at distances up to 300 meters in aerial footage, and handles Indian plate formats including white, yellow, and green registration plates. Night mode with IR-enhanced detection is also supported for 24/7 operation.

Applications

Deployed for traffic management, parking enforcement, and law enforcement surveillance use cases. The system integrates with the ASTROC Sentinel GCS to overlay plate data on the mission map in real-time.