摘要
针对公路上交通压力越来越大、汽车管理的效率越来越低的问题,提出了一种对汽车车牌进行智能数字识别的方法。该方法利用离散型Hopfield神经网络(DHNN)的联想记忆功能,首先,将具有完整信息的数字点阵输入到离散型Hopfield神经网络中,对网络进行训练;其次,用旋转、遮挡和施加高斯白噪声来模拟现实中汽车车牌在识别过程中所遇到的干扰;最后,将这些受到干扰影响而残缺不全的信息输入到神经网络中,让网络进行联想记忆。这三种数字识别的仿真结果表明,离散型Hopfield神经网络可以很好地将信息还原,并且收敛速度很快。
Focused on the increasing traffic pressure on the highway, and the decling efficiency of car management, an intelligent figure recognition method of car license plate was proposed by using the associative memory function of Discrete Hopfield Neural Network（ DHNN）. Firstly, the figure lattice with complete information was input into the DHNN, and the network was trained; secondly, the interference encountered in the process of identifying the real car license plate was simulated by using rotation, covering and Gaussian white noise; finally, the incomplete information affected by interference was input into the neural network, and it was associated and memorized by the network. Simulation results of these three figure identifications show that DHNN can restore information very well, in addition, convergence speed of DHNN is very fast.
出处
《计算机应用》
CSCD
北大核心
2017年第A01期187-189,共3页
journal of Computer Applications
基金
国家自然科学基金资助项目(61174025)
四川省科技计划项目(2014FZ0084)
关键词
数字识别
HOPFIELD神经网络
高斯白噪声
联想记忆
信息还原
figure identification
Hopfield neural network
Gaussian white noise
associative memory
information restoration
