基于卷积神经网络LeNet-5的货运列车车号识别研究  

Research on freight train license recognition based on convolutional neural network LeNet-5

作　　者：王晓锋[1] 马钟[2] 

机构地区：[1]山西医科大学汾阳学院,山西汾阳032200 [2]西北工业大学计算机学院,陕西西安710129

出　　处：《现代电子技术》2016年第13期63-66,71,共5页Modern Electronics Technique

基　　金：国家自然科学基金(61171156)支持项目

摘　　要：针对货运列车车号字符识别,提出了基于卷积神经网络Le Net-5的改进识别方法,考虑到卷积神经网络的层次化以及局部领域等结构特点,对网络中各层特征图的数量及大小等参数进行相应的改进,形成了适用于货运车号识别的新网络模型。实验结果表明,该方法对车号的断裂、污损等问题的解决有较强的鲁棒性,达到了较高的识别率,为整个车号识别系统的精确性提供了保障。For the character recognition of freight train license,the improved recognition method based on convolutional neu?ral network LeNet?5 is proposed. Considering the structural features of the hierarchical convolutional neural network and local field,the parameters of quantity and size of each layer feature pattern in the network were improved correspondingly to form the new network model suitable for the freight train license recognition. The experimental results show that the proposed method has strong robustness to solve the license breakage and stain,and high recognition rate,which provides a guarantee for the accuracy of the entire license recognition system.

关 键 词：列车车号 车号识别 卷积神经网络 LeNet-5 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]