Automatic License Plate Detection Solution

SUMMARY

A global video analytics company partnered with Folio3 AI to develop an Automatic Number Plate Recognition solution for smarter, safer city surveillance.

The goal: to automatically identify vehicle number plates across multiple visual sources while supporting both printed and handwritten plates in authority-approved formats.

Folio3 AI built a computer vision and OCR-based solution capable of processing images, recorded videos, and live camera streams, with on-site deployment for integration into the client's existing surveillance environment.

ABOUT THE CUSTOMER

The client is a global company specializing in video content analytics platforms designed to improve urban intelligence and public safety through advanced surveillance technology.

Their surveillance ecosystem required reliable vehicle identification capabilities that could operate across different visual input sources and integrate with their existing infrastructure.

The organization partnered with Folio3 AI to add automated license plate recognition to its broader video analytics capabilities.

THE CHALLENGE

Urban surveillance environments generate a continuous flow of visual data from fixed cameras, recorded footage, and uploaded images. Manually reviewing this volume of content to identify vehicle license plates is time-consuming, difficult to scale, and highly dependent on operator attention.

The client needed an automated license plate recognition solution that could work reliably across different input sources, recognize multiple plate formats, and integrate with the surveillance infrastructure already in place. The system also had to perform consistently despite variations in image quality, camera position, lighting, motion, and plate appearance.

Three key technical challenges shaped the project:

• Multi-Source Plate Detection: The solution needed to process license plates from static images, recorded videos, and live camera feeds within a single workflow. Each source introduces different technical conditions, including changing frame rates, image resolution, motion blur, viewing angles, and lighting.
• Printed & Handwritten Recognition: The recognition engine had to support both printed and handwritten number plates based on standardized formats approved by local authorities. This added complexity because handwritten characters can vary significantly in shape, spacing, alignment, and clarity.
• On-Site Deployment: The client required the solution to operate within its existing surveillance environment rather than relying entirely on externally hosted processing. This meant the architecture had to be suitable for on-site deployment, integrate with existing camera and surveillance workflows, and operate within the client's available infrastructure.

OUR DELIVERY APPROACH

Folio3 AI developed an end-to-end ANPR solution combining computer vision, machine learning, and OCR capabilities to detect vehicle number plates and convert visual plate information into readable data.

The solution was designed around the client's surveillance workflows, allowing different visual sources to be submitted through a frontend application and processed through the recognition pipeline.

Deployment was completed on-site so the system could integrate with the client's existing surveillance infrastructure.

TOOLS & TECHNOLOGIES

THE SOLUTION

Folio3 AI designed and developed a comprehensive Automatic Number Plate Recognition system that acts as an intelligent vehicle-identification layer within the client's surveillance environment.

The solution accepts images, recorded videos, and live camera streams as input. Computer vision algorithms locate vehicle number plates within incoming footage, while the recognition layer extracts plate information for further use within the surveillance platform.

The application was designed to recognize both printed and handwritten plates using the standardized formats required by local authorities.

A frontend interface also allowed users to upload visual inputs and manage camera sources without interacting directly with the underlying AI pipeline.

The Four-Stage ANPR Pipeline

Folio3 AI developed a four-stage workflow that converts raw surveillance footage into recognized vehicle number plate information.

• Visual Input Ingestion: The system receives images, recorded videos, or live camera feeds through the application and prepares the incoming visual data for automated processing.
• Vehicle & Plate Detection: Computer vision models analyze each relevant frame to identify vehicles and locate the number plate regions required for further recognition.
• Plate Recognition: Detected plate regions are processed through OCR and machine learning models to recognize characters from both printed and handwritten number plates.
• Surveillance System Integration: Recognized plate information is made available through the application, while the on-site deployment enables integration with the client's existing surveillance infrastructure.

SOLUTION ARCHITECTURE

RESULTS ACROSS SMART CITY SURVEILLANCE WORKFLOWS