摘要
车牌识别系统是智能交通管理的重要部分,而车牌字符识别是智能交通系统的核心内容,目前,传统的浅度学习神经网络BP(Back Propagation)人工神经网络因其优越的性能而广泛应用到车牌识别中,但是BP神经网络在局部极值、假饱和、收敛速度缓慢等方面存在着不足。而深度学习[1]与浅度学习相比,其网络结构更接近实际的生物神经网络,因此具有更强的能力,可以很好地提高车牌字符的识别率。其中深度学习神经网络中的卷积神经网络,在视觉图像处理领域进行的实验,得到了很好的结果。因此,将卷积神经网络应用于车牌字符的识别。实验得出:相比于BP神经网络,卷积神经网络的识别率提高了大约5%。
The license plate character recognition is the core of intelligent transportation system.This paper is the convolution neural network to identify the license plate characters,the main method is by changing the traditional convolution of a neural network[2]output unit number and increase the number of convolution layer C5 feature map to achieve better recognition results.The experiment results show that the convolutional neural network recognition rate reached 98.5%,compared to the BP neural network,an increase of approximately 5%.
出处
《工业控制计算机》
2017年第3期51-52,共2页
Industrial Control Computer
关键词
车牌识别
字符识别
深度学习
卷积神经网络
license plate recognition
character recognition
deep learning
convolutional neural network