Defense Technology of License Plate Recognition System Based on Regularization and Image Denoising

Haitong Shen

2024

Abstract

The license plate recognition (LPR) system plays an important role in modern traffic, but the existing models based on deep learning are difficult to deal with in a series of complex situations, including adversarial example attacks. Based on this, this project deeply studies the defense technology of the LPR system utilizing image processing techniques, aiming to enhance the LPR system's processing ability in the face of complex environments, especially adversarial sample attacks. In terms of research methods, this project is based on a GRU-based sequence model, using regularization and image noise reduction to weaken the sensitivity of the model to specific perturbed data points and enhance the robustness of the model, so that the accuracy of the model to identify adversarial samples is improved from 22% to 68%. The experimental results show that the regularization and image noise reduction method can make the LPR model effectively distinguish the original license plate from the adversarial samples, and can restore the original information in the image adversarial samples. There are still some problems in the model, such as partial errors that may occur when trying to restore the authentic data encoded on the license plate, especially the abbreviation of the province. Further improvement of the character recognition method can be considered in the future.

Paper Citation

in Harvard Style

Shen H. (2024). Defense Technology of License Plate Recognition System Based on Regularization and Image Denoising. In Proceedings of the 2nd International Conference on Data Analysis and Machine Learning - Volume 1: DAML; ISBN 978-989-758-754-2, SciTePress, pages 329-333. DOI: 10.5220/0013516600004619

in Bibtex Style

@conference{daml24,
author={Haitong Shen},
title={Defense Technology of License Plate Recognition System Based on Regularization and Image Denoising},
booktitle={Proceedings of the 2nd International Conference on Data Analysis and Machine Learning - Volume 1: DAML},
year={2024},
pages={329-333},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013516600004619},
isbn={978-989-758-754-2},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 2nd International Conference on Data Analysis and Machine Learning - Volume 1: DAML
TI - Defense Technology of License Plate Recognition System Based on Regularization and Image Denoising
SN - 978-989-758-754-2
AU - Shen H.
PY - 2024
SP - 329
EP - 333
DO - 10.5220/0013516600004619
PB - SciTePress