Research Article
Chinese License Plate Recognition System Design Based on YOLOv4 and CRNN + CTC Algorithm
@INPROCEEDINGS{10.1007/978-3-030-89814-4_63, author={Le Zhou and Wenji Dai and Gang Zhang and Hua Lou and Jie Yang}, title={Chinese License Plate Recognition System Design Based on YOLOv4 and CRNN + CTC Algorithm}, proceedings={Mobile Multimedia Communications. 14th EAI International Conference, Mobimedia 2021, Virtual Event, July 23-25, 2021, Proceedings}, proceedings_a={MOBIMEDIA}, year={2021}, month={11}, keywords={License plate recognition YOLOv4 Convolutional recurrent neural network Connectionist temporal classification}, doi={10.1007/978-3-030-89814-4_63} }- Le Zhou
Wenji Dai
Gang Zhang
Hua Lou
Jie Yang
Year: 2021
Chinese License Plate Recognition System Design Based on YOLOv4 and CRNN + CTC Algorithm
MOBIMEDIA
Springer
DOI: 10.1007/978-3-030-89814-4_63
Abstract
License plate recognition (LPR) is widely used in the intelligent transportation systems. Traditional recognition methods have many disadvantages with slow detection speed and low recognition accuracy. In order to solve these problems, this paper proposes an end-to-end LPR method, which is based on YOLOv4 and Convolutional Recurrent Neural Network (CRNN) with Connectionist Temporal Classification (CTC) algorithm, which can effectively improve the detection speed and accuracy. First, based on the excellent classification and detection performance of YOLOv4, it is applied to accurately locate the license plate of the input car image. Then, we use CRNN to recognize the character information imported in the license plate image and add the CTC algorithm to the CRNN network to achieve the alignment of the input and output formats of the character information. Experimental results show that the accuracy rate of license plate recognition detection reaches as high as 97%, and the detection speed is as low as around 30 FPS (Frames Per Second).
