A method, system, medium and terminal based on license plate character recognition

Abstract

The present invention provides a method, system, medium and terminal based on license plate character recognition. The method includes: acquiring license plate image information; establishing a license plate recognition network based on the license plate image information, and the network includes a license plate row recognition model and a license plate Chinese character recognition model ; Input the image information of the license plate to be recognized into the trained model respectively, and obtain the result of the whole row of the license plate and the result of the Chinese characters on the license plate; Obtain the final license plate character recognition result; the present invention is aimed at the license plate recognition task, by simultaneously extracting the features of the whole line of license plate and the Chinese character area of ​​the license plate, it can extract two features for the difficulty of Chinese character recognition in the license plate character recognition, and obtain the result of the Chinese character of the license plate twice , by configuring the default Chinese characters of the province and correcting the Chinese characters of the recognition results, the recognition rate of the license plate is further improved.

Description

technical field [0001] The invention relates to the field of computer applications, in particular to a license plate character recognition method, system, medium and terminal. Background technique [0002] With the continuous development of computer vision, the recognition of license plate numbers of vehicles in traffic through character recognition technology has been widely used in the field of intelligent transportation. However, due to recognition scenarios such as weather, defacement, occlusion, and large angles, it poses great challenges to license plate character recognition. License plate character recognition generally adopts four methods, namely based on template matching method, based on neural network method, method based on support vector machine and method based on deep learning. The first three methods are called traditional methods. The traditional method needs to manually select features. The robustness of the traditional method is poor. Deep learning has b...

AI Technical Summary

Problems solved by technology

The first three methods are called traditional methods. The traditional method needs to manually select features. The robustness of the traditional method is poor. Deep learning has become popular in recent years. The method based on deep learning will automatically extract features. When the training samples are relatively When rich, the recognition results of the network tend to be better

[0003] However, the current recognition methods either segment the license plate characters and recognize each character individually, or use a deep learning network to automatically segment and recognize the license plate characters, or use multiple networks to recognize the license plate characters, without considering the license plate characters. Focus on learning difficult characters

Images