摘要
提出一种基于整数比例算法识别和卷积神经网络(convolutional neural network,CNN)相结合的高效高精度车牌识别系统,通过图像预处理模块对图像进行二值化、切割和缩放,通过坐标生成模块进行坐标检测,采用CNN对预处理的图片进行识别.实验结果表明,该系统根据车牌字符比较规则的特点,缩减了网络模型的复杂度,车牌的识别准确率和识别速度均有所提高.
A high efficiency and high precision license plate recognition system based on the integer ratio algorithm recognition combined with convolutional neural network(CNN)is proposed.The image is binarized,cut and scaled through the image preprocessing module,coordinate detection is completed in the coordinate generation module,and then the image preprocessed is identified by the CNN.The experimental results show that the system reduces the complexity of the network model according to the characteristic that the characters on the license plate are relatively regular,and the recognition accuracy and speed for the license plate are improved.
作者
刘曜溪
徐志强
黄太一
谷文启
李蕴晖
杜高明
LIU Yaoxi;XU Zhiqiang;HUANG Taiyi;GU Wenqi;LI Yunhui;DU Gaoming(School of Electronic Science&Applied Physics,Hefei University of Technology,Hefei 230601,China)
出处
《中国科技论文在线精品论文》
2020年第4期456-464,共9页
Highlights of Sciencepaper Online
基金
合肥工业大学大学生创新创业训练计划项目(S201910359255)
关键词
电子技术
车牌识别系统
硬件加速器
卷积神经网络
FPGA
electronic technology
license plate recognition system
hardware accelerator
convolutional neural network
FPGA
作者简介
刘曜溪(1999—),男,本科生,主要研究方向:微电子;通信联系人:杜高明,副教授,主要研究方向:多核体系结构、片上网络体系结构、多核、NoC性能分析.E-mail:dugaoming@hfut.edu.cn
