CAJ | 학술논문

根据我国车辆牌照的特点,提出了基于BP神经网络识别算法。算法中将分类器分为汉字分类器,英文字母分类器,英文字母和数字混合分类器以及数字分类器四种,这种神经网络设计可以有效简化网络结构,提高识别精度和速度。由于标准BP算法具有收敛速度慢、易陷入局部极小点等缺点,对BP算法进行了改进。通过仿真实验,该字符识别系统具有较高的识别率,同时也具备了神经网络本身容错能力强,即鲁棒性好的特点。
근거아국차량패조적특점,제출료기우BP신경망락식별산법。산법중장분류기분위한자분류기,영문자모분류기,영문자모화수자혼합분류기이급수자분류기사충,저충신경망락설계가이유효간화망락결구,제고식별정도화속도。유우표준BP산법구유수렴속도만、역함입국부겁소점등결점,대BP산법진행료개진。통과방진실험,해자부식별계통구유교고적식별솔,동시야구비료신경망락본신용착능력강,즉로봉성호적특점。
According to the vehicle licence in China characteristics,paper presented recognition based on BP neural network algorithm.Classifier in algorithm was divided into Chinese characters,English letter classification,English letters and numbers mixed classification,as well as for digital classification for four.The neural network structure design can effectively simplify network,improve recognition accuracy and speed.Due to the slow convergence standard BP algorithm,ease into a local minimum point disadvantage,paper improved BP algorithms.Through simulation,the character recognition system had a high recognition rate,also had a neural network fault-tolerant ability in itself,features of robustness was good.