Robust License Plate Recognition in OCC-Based Vehicle Networks Using Image Reconstruction

Area of Science:

• Computer Vision
• Internet of Vehicles (IoV)
• Optical Camera Communication (OCC)

Background:

• Optical Camera Communication (OCC) enables vehicle-to-infrastructure communication using traffic lights and street cameras.
• OCC light emissions create stripe patterns in images, degrading the accuracy of license-plate recognition systems.
• Existing license-plate recognition systems struggle with interference from OCC in Internet of Vehicles (IoV) applications.

Purpose of the Study:

• To propose and demonstrate a method for reducing OCC stripe interference in license-plate recognition.
• To enhance the accuracy of license-plate recognition in environments utilizing Optical Camera Communication.

Main Methods:

• Developed an innovative pipeline featuring an end-to-end image reconstruction module.
• Modeled OCC stripes as multiplicative noise to synthesize a pairwise dataset from existing license-plate data.
• Built a prototype to simulate real-world OCC-based vehicle network scenarios and collect data.

Main Results:

• Achieved 81.58% recognition performance on a synthesized dataset and 79.35% on real-scene captured data.
• Demonstrated significant improvements of approximately 31.18% and 24.26% compared to conventional methods.
• The proposed method effectively reconstructs high-quality license-plate images for recognition under OCC conditions.

Conclusions:

• The developed method successfully mitigates OCC stripe interference, enhancing license-plate recognition accuracy.
• The approach provides a viable solution for reliable vehicle identification in OCC-enabled IoV systems.
• This research contributes to the practical deployment of Optical Camera Communication in intelligent transportation systems.