摘要
针对现有的车牌识别系统在遇到复杂条件,例如暗光、遮挡、多车牌、能见度低等情况时,难以有效地定位并识别车牌,提出了一种基于卷积神经网络的车牌自动识别系统.在车牌定位阶段综合应用3种定位方式对车牌进行初步定位检测,然后使用CNN模型对检测到的候选车牌进行判断;在车牌字符识别阶段,将分割出的字符输入到设计好的卷积神经网络模型中进行训练,得到的输出结果即为识别的车牌字符.在5906张车牌图像和非车牌图像以及36261张字符图片上的实验结果表明:提出的车牌识别系统对车牌和字符的识别率分别达到了94%和96.4%,明显优于传统的车牌识别方法,具有极高的实用性,可以满足绝大多数场景的使用需求.
The existing license plate recognition system is difficult to locate and identify the license plate effectively when it encounters complex conditions such as dim light,plate is blocked,multiple plates and low visibility. An automatic license plate recognition system based on convolution neural network( CNN) has been proposed in this paper. In the license plate location phase,three kinds of positioning methods are integrated for the initial locating of the license plate. Then,the CNN model is used to judge the selected license plate. In the license plate character recognition phase,segmented characters are input to a designed CNN model,and the output of the CNN model is the result of the recognized characters. The experiment is based on 5906 license plate images and non-license plate images,and 36261 characters images. The results of the experiment show that the recognition rates of the proposed system for license plate and character are 94% and 96. 4% respectively,which is significantly better than that of traditional license plate recognition methods. It meets the needs of the vast majority of the use of the scene,with a high practicality.
出处
《中南民族大学学报（自然科学版）》
CAS
北大核心
2017年第3期125-130,共6页
Journal of South-Central University for Nationalities：Natural Science Edition
基金
国家自然科学基金资助项目(61302192)
关键词
车牌定位
车牌识别
字符识别
卷积神经网络
license plate locating
license plate recognize
character recognize
CNN
