An Improved LeNet-5 Convolutional Neural Network for Intelligent Recognition of License Plate Images

Abstract

In intelligent transportation systems, accurate license plate recognition is an important component. This paper briefly introduces the LeNet-5 model for license plate image recognition. We improved the model by introducing an inception-SE convolution module. In simulation experiments, the optimized LeNet-5 model was compared with the original LeNet-5 model and a back-propagation neural network (BPNN). The results showed that the characters after preprocessing and character segmentation were clearer than those in the original images. During training, the optimized LeNet-5 converged the fastest, reached stability after 100 iterations, and had the smallest error after stability. The overall recognition accuracy of the BPNN model for the license images was 64.3%. For the original LeNet-5 model, it was 84.0%, and for the optimized LeNet-5 model, it was 98.6%.