A method for license plate recognition in low-resolution conditions

School of Information and Intelligent Engineering, Zhejiang Wanli University, 315000, Ningbo, China

^* Corresponding author: 2022014439@zwu.edu.cn

Abstract

With the widespread application of license plate recognition technology, the development of license plate recognition technology is becoming increasingly advanced. However, in some low-resolution situations, the efficiency and accuracy of license plate recognition are greatly reduced. To understand the development in this area, this paper conducts research from two perspectives: traditional methods and deep learning methods. Among the traditional methods, this paper discusses methods based on ALBP, methods based on image vertical projection feature analysis and geometric features of character segmentation algorithms, and methods based on character region perception of end-to-end license plate recognition algorithms. Among the license plate recognition methods based on deep learning, this paper discusses algorithms such as AOD-Net, YOLOv5, LPRNet, CRNN+CTC, and FasterR-CNN. The purpose of this review is to help readers understand how researchers have overcome low resolution by improving the existing recognition methods in the development of license plate recognition technology, and point out the shortcomings of each improved method so that readers can understand the development and existing problems in this field.