Edge‑Optimized Real‑Time Deep Convolutional Framework for Robust Multilingual Vehicle License Plate Detection and Recognition Under Diverse Environmental Conditions

Om Prakash Yadav, Sabbina SuryaSaranya, Nandhini S, R Rajkumar, Lokasani Bhanuprakash, Karthik P

2025

Abstract

The network presents an edge-optimized deep convolutional architecture that provides realtime highaccuracy detection and multilingual recognition of vehicle license plates under different lighting, weather conditions. With the combination of geometric normalization, motionaware ROI tracking and adaptive illumination correction, the system is resistant to skew, occlusion, lowresolution and can also provide a highaccuracy analysis. A lightweight, pruned/quantized backbone for embedded GPUs supports >30 FPS and offers explainable output through GradCAM and SHAP visualizations. The model is equipped with the continuous online learning and domainadaptation modules to effectively update itself over time and maintain robustness in a global plate style manner. We present extensive benchmarking against the variants of ResNet, EfficientNet and MobileNet using the new 2021-2025 heterogeneous dataset to justify our efficacy in terms of both accuracy and latency.

Paper Citation

in Harvard Style

Yadav O., SuryaSaranya S., S N., Rajkumar R., Bhanuprakash L. and P K. (2025). Edge‑Optimized Real‑Time Deep Convolutional Framework for Robust Multilingual Vehicle License Plate Detection and Recognition Under Diverse Environmental Conditions. In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies - Volume 1: ICRDICCT`25; ISBN 978-989-758-777-1, SciTePress, pages 336-341. DOI: 10.5220/0013863600004919

in Bibtex Style

@conference{icrdicct`2525,
author={Om Prakash Yadav and Sabbina SuryaSaranya and Nandhini S and R Rajkumar and Lokasani Bhanuprakash and Karthik P},
title={Edge‑Optimized Real‑Time Deep Convolutional Framework for Robust Multilingual Vehicle License Plate Detection and Recognition Under Diverse Environmental Conditions},
booktitle={Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies - Volume 1: ICRDICCT`25},
year={2025},
pages={336-341},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013863600004919},
isbn={978-989-758-777-1},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies - Volume 1: ICRDICCT`25
TI - Edge‑Optimized Real‑Time Deep Convolutional Framework for Robust Multilingual Vehicle License Plate Detection and Recognition Under Diverse Environmental Conditions
SN - 978-989-758-777-1
AU - Yadav O.
AU - SuryaSaranya S.
AU - S N.
AU - Rajkumar R.
AU - Bhanuprakash L.
AU - P K.
PY - 2025
SP - 336
EP - 341
DO - 10.5220/0013863600004919
PB - SciTePress