基于卷积神经网络的车牌识别研究
License Plate Recognition Based on Convolutional Neural Networks
齐佳鑫 1张志华 1付金尉 1贺紫菡 1司志广 1王艺雄1
作者信息
- 1. 辽宁科技大学电子与信息工程学院, 辽宁 鞍山 114051
- 折叠
摘要
为了满足交通管理和安全监控需求,针对车牌识别技术在智能交通系统中的重要性,本文旨在运用卷积神经网络实现汽车牌照的识别,首先通过对采集到的图像进行滤波降噪等预处理,运用Sobel算子进行边缘检测;其次对图像进行形态学运算,利用轮廓检测确定出车牌的位置;然后采取边缘检测的方法进行字符分割;最后运用卷积神经网络对车牌字符实现识别.模型测试结果表明,车牌识别系统汉字识别的准确率为 91.4%,数字和字母共同识别的准确率为 95.5%.
Abstract
In order to meet the demand for traffic management and safety monitoring,aiming at the important role of license plate recognition technology in the intelligent transportation system.This paper aims to use convolutional neural networks to recognize car license plates.Firstly,pre-processing such as filtering and denoising is performed on the collected images,and Sobel operators are used for edge detection;Secondly,perform morphological operations on the image and use contour detection to determine the position of the license plate;then,edge detection method is adopted for character segmentation;finally,convolutional neural networks are used to recognize license plate characters.The model test results show that the accuracy of Chinese character recognition in the license plate recognition system is 91.4%,and the accuracy of jointly recognizing numbers and letters is 95.5%.
关键词
神经网络/模型/车牌识别/车牌识别系统/边缘检测Key words
neural network/model/license plate recognition/license plate recognition system/edge detection引用本文复制引用
基金项目
辽宁科技大学大学生创新创业训练计划(X202310146303)
出版年
2024
NETL
NSTL
万方数据