基于卷积神经网络的免字符分割车牌识别算法

License plate recognition algorithm without character segmentation based on convolutional neural network

随着智能停车场和自动收费系统的普及,智能交通管理系统在车牌识别方面对速度和精度提出了更高的要求。为了提高车牌识别的泛化能力,本文提出了一种基于改进的CRNN+CTC免字符分割车牌字符识别算法,在现有的卷积神经网络的CRNN+CTC网络模型基础上分别对CNN和RNN进行替换后引入CTC函数,并有针对性地增加训练数据集,在训练过程中使用了批量归一化算法来加快学习速度。实验结果表明,该算法的鲁棒性与通用性均有一定提升,在复杂环境下车牌识别的准确性更高,达到了改进效果。

With the construction of intelligent parking lot and urban traffic and the popularization of automatic toll collection system,intelligent traffic management system puts forward higher requirements for speed and accuracy in license plate recognition.In order to improve the generalization ability of license plate recognition,this paper proposes a character segmentation free license plate character recognition algorithm based on improved CRNN+CTC.Based on the existing CRNN+CTC network model,CNN and RNN are replaced respectively,CTC function is introduced,and the training data set is added.Finally,batch normalization algorithm is used in the training process to speed up the learning speed.The experimental results show that the algorithm has higher accuracy in license plate recognition under complex environment,and its robustness and generality are improved.