Research on License Plate Recognition Algorithm Based on ABCNet

Abstract

In order to improve the accuracy of license plate recognition algorithm, we propose a license plate recognition algorithm based on ABCNet. Firstly, the original images with ABCNet is to locate license plate detection network, secondly, use CRNN - CTC algorithm for license plate character recognition, character recognition algorithm is the main process of the convolution neural network is used to extract image convolution, the cycle of neural network is used to extract image convolution features of sequence, the CTC algorithm is used to solve the problem that training characters cannot be aligned, the license plate recognition's accuracy is 96.4%. With a speed of 9ms, it has a well effect and can play an important role in actual traffic management.