摘要
车牌识别是智能交通系统中一个重要的环节,它可以应用到很多领域,如高速公路自动收费、交通监控系统、停车场管理等。提出一种改进的LM-BP神经网络车牌字符识别方法,该方法根据国内现行车牌编制的特点,结合LM算法改进传统BP神经网络,并增加σ参数修正Ⅲ-BP算法,避免传统BP神经网络收敛速度缓慢并容易陷入局部极小值的缺点,进行了大量实验,达到了预期的识别效果和收敛速度。
As the important part of intelligent transportation system,license plate recognition system is of great significance for many fields,such as automated highway toll collection,traffic monitoring systems,parking management.An improved LM-BP neural network recognition method is presented.Based on the characteristics of the current Chinese domestic plate preparation,it combines LM algorithm and σ parameter to improve the traditional LM BP neural network algorithm,to avoid the traditional BP neural network convergence speed being slow and prone to fall into local minima.A lot of experiments are conducted to achieve the desired effect of recognition and convergence speed.
出处
《测控技术》
CSCD
2016年第2期48-51,57,共5页
Measurement & Control Technology
关键词
智能交通系统
车牌字符识别
LM-BP神经网络
intelligent transportation system
license plate character recognition
LM-BP neural network