A Comparative Study of OCR Architectures for Korean License Plate Recognition: CNN–RNN-Based Models and MobileNetV3–Transformer-Based Models
Sensors · 12.02.2026 · 26 (4) · 1208
DOI: 10.3390/s26041208 ↗Kurz gefasst
Fachartikel in der Zeitschrift Sensors, online erschienen am 12.02.2026. Verfasst von Seungju Lee und 1 weiteren Autor:innen. Beteiligte Einrichtung: Department of Smart ICT Convergence Engineering, Seoul National University of Science and Technology, Seoul 01811, Republic of Korea. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 26 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Institute of Information & Communications Technology Planning & Evaluation.
Abstract (Originalsprache)
This paper presents a systematic comparative study of optical character recognition (OCR) architectures for Korean license plate recognition under identical detection conditions. Although recent automatic license plate recognition (ALPR) systems increasingly adopt Transformer-based decoders, it remains unclear whether performance differences arise primarily from sequence modeling strategies or from backbone feature representations. To address this issue, we employ a unified YOLOv12-based license plate detector and evaluate multiple OCR configurations, including a CNN with an Attention-LSTM decoder and a MobileNetV3 with a Transformer decoder. To ensure a fair comparison, a controlled ablation study is conducted in which the CNN backbone is fixed to ResNet-18 while varying only the sequence decoder. Experiments are performed on both static image datasets and tracking-based sequential datasets, assessing recognition accuracy, error characteristics, and processing speed across GPU and embedded platforms. The results demonstrate that the effectiveness of sequence decoders is highly dataset-dependent and strongly influenced by feature quality and region-of-interest (ROI) stability. Quantitative analysis further shows that tracking-induced error accumulation dominates OCR performance in sequential recognition scenarios. Moreover, Korean license plate–specific error patterns reveal failure modes not captured by generic OCR benchmarks. Finally, experiments on embedded platforms indicate that Transformer-based OCR models introduce significant computational and memory overhead, limiting their suitability for real-time deployment. These findings suggest that robust license plate recognition requires joint consideration of detection, tracking, and recognition rather than isolated optimization of OCR architectures.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Department of Smart ICT Convergence Engineering, Seoul National University of Science and Technology, Seoul 01811, Republic of Korea.
ORCID-Profile: Seungju Lee, Gooman Park
Förderung
- Institute of Information & Communications Technology Planning & Evaluation
Zitieren
APA
Lee, S., & Park, G. (2026). A Comparative Study of OCR Architectures for Korean License Plate Recognition: CNN–RNN-Based Models and MobileNetV3–Transformer-Based Models. Sensors, 26(4), 1208. https://doi.org/10.3390/s26041208
Vancouver
Lee S, Park G. A Comparative Study of OCR Architectures for Korean License Plate Recognition: CNN–RNN-Based Models and MobileNetV3–Transformer-Based Models. Sensors. 2026;26(4):1208. doi:10.3390/s26041208
BibTeX
@article{Lee2026,
author = {Lee, Seungju and Park, Gooman},
title = {A Comparative Study of OCR Architectures for Korean License Plate Recognition: CNN–RNN-Based Models and MobileNetV3–Transformer-Based Models},
journal = {Sensors},
year = {2026},
volume = {26},
number = {4},
pages = {1208},
doi = {10.3390/s26041208}
}
EUVIMED