摘要
车牌识别在高速收费口、小区车辆出入口、停车场自动收费系统等方面得到越来越多的应用,这在一定程度上可以减少交通道路的拥挤,缓解交通压力。本文应用OpenCV库相关功能完成车牌的定位以及字符的分割,在此基础上利用TensorFlow框架的Keras模块搭建卷积神经网络,对车牌中的汉字、数字和字母分别进行识别,其中车牌汉字模型评估的准确率为92.4%,数字和字母一起识别模型评估的准确率为95.6%,识别效果较好。
License plate recognition is being used more and more widely in high-speed toll gates, community vehicle entrances and exits, parking lot automatic toll collection systems, etc., which can reduce traffic congestion to a certain extent and relieve traffic pressure. This paper proposes to use the functions of the OpenCV library to complete license plate positioning and the segmentation of the characters. On this basis, a convolutional neural network is built by using Keras module of TensorFlow framework for recognizing the Chinese characters, numbers and letters in the license plate. The accuracy rate of the license plate Chinese character model evaluation is 92.4%, and the accuracy rate of model evaluation for identifying numbers and letters together is 95.6%, which verifies a good recognition effect.
作者
高艳
刘海峰
GAO Yan;LIU Haifeng(Jinzhong College of Information,Taigu 030800,China)
出处
《软件工程》
2022年第5期23-25,22,共4页
Software Engineering
基金
山西省高等学校科技创新项目(2020L0745)。
关键词
车牌定位
字符分割
卷积神经网络
车牌识别
license plate positioning
character segmentation
convolutional neural network
license plate recognition
