基于SCG-BP神经网络的车牌字符识别  

License Plate Character Recognition Based on SCG-BP Neural Network

作　　者：李非 LI Fei(School of Electronic Science,Northeast Petroleum University,Daqing 163000)

机构地区：[1]东北石油大学电子科学学院,大庆163000

出　　处：《计算机与数字工程》2021年第6期1229-1233,1252,共6页Computer & Digital Engineering

摘　　要：针对传统BP网络算法存在车牌字符识别速度慢和准确率低的问题,提出了一种SCG优化的BP神经网络车牌字符识别的算法。通过对BP神经网络的输入和算法进行改进实现提高神经网络对字符的识别效率。对输入的优化是使用主成分分析法进行车牌字符特征提取,将提取的特征作为BP神经网络的输入。对算法的优化是使用成比例共轭梯度下降法寻找网络最优连接权重。仿真实验表明,SCG-BP神经网络大幅度缩短识别时间并且提高了准确率,确定隐含层神经元个数为110。该算法对车牌字符的识别率可以达到95%以上,取得结果达到预期,改进的算法有一定的实践可行性。Aiming at the problems of slow speed and low accuracy of license plate character recognition in traditional BP net⁃work algorithm,a license plate character recognition algorithm based on SCG optimized BP neural network is proposed.The input and algorithm of BP neural network are improved to improve the character recognition efficiency of neural network.To optimize the input,principal component analysis is used to extract the license plate character features,and the extracted features are taken as the input of BP neural network.The algorithm is optimized by using proportional conjugate gradient descent to find the optimal con⁃nection weight of the network.Simulation experiments show that SCG-BP neural network greatly reduces the recognition time and improves the accuracy,and the number of hidden layer neurons is determined to be 110.The recognition rate of the license plate character can reach more than 95%,and the result reaches the expectation.The improved algorithm has certain practical feasibility.

关 键 词：车牌字符识别 主成分分析法 SCG BP神经网络 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]