Work

A Comparative Study on Thai License Plate Recognition: Object Detection and Transformer Learning Approaches

Year: 2025

Type: article

Abstract: Car theft remains a significant issue in Thailand, necessitating advanced security solutions. This research investigates the application of License Plate Recognition (LPR) technology using deep learning models for vehicle and license plate detection, as well as both YOLO and Transformer models for license plate reading. Experimental results demonstrate that YOLOv5nu achieved a high Mean Average Precision (mAP) of 0.995 for license plate detection, while the Transformer model excelled in license plate reading with an accuracy of 0.9108 and a loss of 0.0326. Despite these promising results, the models faced limitations in low-light conditions and complex plate layouts, affecting detection accuracy. Future work should focus on enhancing the models’ adaptability to real-world environments, expanding datasets to include diverse scenarios, and improving robustness against environmental variations. These findings underscore the potential of AI-driven LPR systems in advancing vehicle security technology.

Source: IEEE Access

DOI: 10.1109/access.2025.3575811

Authors: Kwankamon Dittakan, Jirawat Thaenthong, Thanakorn Prasomkit

Institution: Prince of Songkla University

Language: English

Topic: Vehicle License Plate Recognition

Subfield: Media Technology

Field: Engineering

Domain: Physical Sciences

SDG: Industry, innovation and infrastructure

Open Access status: gold

Estimated APC paid: $2,075