License plate character global recognition method based on deep learning model

Abstract

The invention provides a license plate character global recognition method based on a deep learning model. The license plate character global recognition method comprises the steps of: step a, performing vehicle detection to obtain a target vehicle, and determining an entire license plate detection region by means of the target vehicle; step b, performing detection in the entire license plate detection region to obtain a license plate region; step c, and performing global recognition on the license plate region by means of a deep neural network model, wherein the deep neural network model comprises four convolutional layers and two full-connection layers which are connected in sequence, the last full-connection layer comprises seven branches, and the branches correspond to Chinese characters, letters, numbers or identifiers on license plate numbers separately. The license plate character global recognition method improves the license plate character recognition rate through introducing the deep learning model for global recognition, and overcomes the error recognition problem caused by the dependence and result uncertainty when a traditional license plate recognition method needs to segment license plate characters.

Description

technical field [0001] The invention relates to the fields of intelligent transportation systems and digital security, in particular to a method for overall recognition of license plate characters based on a deep learning model. Background technique [0002] License plate recognition is a basic part and an important link in the field of intelligent transportation system (ITS). It is necessary for other parts of the intelligent transportation system, such as vehicle logo recognition, body color analysis, vehicle model analysis, bayonet detection, electronic police, and other traffic event related modules. How well it is handled makes a big difference. [0003] The traditional license plate recognition system is roughly divided into several sub-modules such as license plate detection, license plate location, character segmentation and character recognition. Among them, the license plate number recognition part, the character segmentation part relies heavily on the previous po...

AI Technical Summary

Problems solved by technology

[0005] 1. The traditional method requires precise positioning and segmentation of the license plate, which is highly dependent and prone to segmentation errors;

[0006] 2. The stability of the traditional method is not good in the case of poor light, low resolution, contamination, large tilt, etc.;

[0007] 3. It is difficult for a single model to guarantee better identification indicators

Images