A Lightweight Convolutional Neural Network for License Plate Character Recognition

Abstract

License plate recognition is a very important component of intelligent parking lot management systems. The accuracy and speed of license plate recognition directly affect the speed of vehicles entering and leaving a parking lot. According to the characteristics of the character composition of a license plate, based on the convolutional neural network model LeNet-5, the license plate character recognition network model is designed. After training to the model, the recognition accuracy of each character in the license plate is higher than 99.97%, and the recognition duration of a single license plate is as fast as 1.31 ms. In this study, the detailed algorithm description of license plate recognition model is given; meanwhile, the optimization method for the neural network model is summarized.