基于深度学习模型LeNet-5-L的车牌识别算法

A License Plate Recognition Algorithm Based on Deep Learning Model LeNet-5-L

摘要:

针对车牌中汉字识别率低和识别速度慢问题,提出一种基于深度学习的车牌识别网络LeNet-5-L,该网络把车牌识别分为两个阶段,运用OpenCV库函数对车牌图像预处理,结合垂直投影分割方法将车牌分割为7个独立字符图像,降低了图像特征提取难度,从而提高车牌中各个的字符识别率和整个车牌识别速度；运用卷积神经网络解决车牌字符识别问题,基于LeNet-L设计一种车牌字符识别网络LeNet-5-L,有效提高车牌中首字符汉字识别率。实验结果表明,该网络对车牌中各个字符的识别准确率均高于99.97%,单个车牌识别时间仅需0.83ms,该方法有效的提高车牌识别的正确率和识别速度。

Abstract:

Aiming at the problems of low recognition rate and slow recognition speed of Chinese characters in license plate, a license plate recognition network Lenet-5-L based on deep learning is proposed, This network divides license plate recognition into two stages, using OpenCV library function to preprocess license plate image, combining with vertical projection segmentation method to divide license plate into seven independent character images, so as to reduce character recognition feature dimension and improve character recognition rate and speed; using volume neural network to solve the problem of license plate character recognition, a license plate character recognition network LeNet-5-L is designed based on LeNet-L can effectively improve the recognition rate of the first Chinese character in the license plate. The experimental results show that the recognition accuracy of the network is higher than 99.97%,and the recognition time of a single license plate is only 0.83ms. This method effectively improves the recognition accuracy and recognition speed.