End-to-end license plate recognition method based on deep learning

Abstract

The invention provides an end-to-end license plate recognition method based on deep learning. The image of the license plate to be detected is output after being processed through the Gaussian mixturemodel algorithm; a convolutional neural network is used for extracting the shared characteristic module, and the image of the license plate to-be-detected is input the trained convolutional neural network algorithm to extract the characteristics, and after being processed by a license plate image primary detection module, the return position information, the regression angle characteristic mapping graph and a corresponding characteristic mapping graph determining whether the license plate is or not are input; the license plate primary detection module is used for analyzing and obtaining a confidence degree score through the corresponding characteristic mapping graph determining whether the license plate is or not to obtain a batch of candidate license plate images; the license plate images are fused by means of a non-maximum value inhibition algorithm, and the image of the real license plate position can be obtained; the image of the real license plate position is input to a license plate content recognition feedback module to recognize the license plate content. The method is applied to a parking lot management system, the parking lot charging efficiency is improved, the licenseplate recognition efficiency is improved, and the cost is saved.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to an end-to-end license plate recognition method based on deep learning. Background technique [0002] The existing license plate detection method is to use the characteristics of the license plate area to judge the license plate, and to segment the license plate area from the entire vehicle image. The license plate itself has many inherent characteristics, which are different in different countries. my country's license plate has the following characteristics that can be used for detection: (1) the background color of the license plate is generally quite different from the color of the vehicle body and the character color; (2) the license plate has a continuous or discontinuous border due to wear; (3) the characters in the license plate There are several, basically arranged horizontally, and there are rich edges in the rectangular area of ​​the license plate, ...

AI Technical Summary

Problems solved by technology

[0008] Aiming at the deficiencies of the prior art, the present invention provides an end-to-end license plate recognition method based on deep learning, which solves the problems of the traditional Adaboost-based license plate detection algorithm, such as too long training time and over-adaptation of weight adjustment.

Images