License Plate Recognition Model Based on CNN+LSTM+CTC

作     者：Hangming Zhang Feng Sun Xiaopu Zhang Lingling Zheng 

作者机构：College of Instrumentation and Electrical Engineering Jilin University Changchun 130062 China The State Key Laboratory of Automotive Simulation and Control Jilin University Changchun 130061 China 

出 版 物：《国际计算机前沿大会会议论文集》 (International Conference of Pioneering Computer Scientists, Engineers and Educators（ICPCSEE）)

年 卷 期：2019年第2期

页      面：675-678页

学科分类：0303[法学-社会学] 03[法学] 

主　　题：License plates Neural network Model LSTM CTC Recognition 

摘      要：With the continuous improvement of the social and economic level, the number of vehicles has exploded in the city, and traditional manual identification license plates have been unable to meet the demand. In this paper, a Convolutional neural network (CNN)-based license plate recognition system is designed. The recognition module uses the CNN+LSTM+CTC model to simplify the convolutional layer structure to adapt to the lightweight training mode. The two-way LSTM structure is used to learn from both sides of the license plate to enhance the end-to-end recognition effect. Compared with the traditional scheme, the CTC loss calculation method eliminates the need for character alignment, streamlines the steps, and improves the recognition accuracy. The experiment shows that the license plate recognition software system designed in this paper has a high recognition accuracy rate of 98.59%.