Deep learning network sharing based license plate identification method

Abstract

The invention relates to a deep learning network sharing based license plate identification method. The method comprises the following steps that 1) an image to be processed is input to a license plate detection deep learning network, different types of license plate threshold graphs are output; 2) the image to be processed is input to a character detecting deep learning network, and a character detecting threshold graph is output; and 3) according to the license plate threshold graphs and the character detecting threshold graph, an optimal license plate type and license plate number identification result is obtained via integrated determination. The method has the advantages that (1) the size of the processing image of the identification method is not limited, and the image can be processed under the condition that the image is not null; (2) license plate identification is carried out by fusing license plate information with character information, false detection can be eliminated easier, the accuracy is increased, and the computing complexity of subsequent determining is reduced; and (3) network structure and parameters of the deep learning network are shared to the largest degree, and the complexity of training and application is reduced.

Description

technical field [0001] The invention relates to the fields of license plate recognition technology and machine learning technology, in particular to a license plate recognition method based on deep learning network sharing. Background technique [0002] Accurate and fast recognition of license plate numbers has very important practical significance. License plate information is one of the unique identifiers of vehicles, and is widely used in various fields that require vehicle monitoring, such as expressway checkpoints, vehicle violation capture, and residential and shopping mall entrance and exit monitoring. The research on license plate recognition method also has certain theoretical value. As a type of license plate, the license plate recognition method can be extended and applied to the recognition of other types of license plates. [0003] As a rapidly emerging machine learning method, deep learning has been widely used in various fields such as computer vision, speec...

AI Technical Summary

Problems solved by technology

[0005] There are still some problems in the existing license plate recognition method based on deep learning: (1) There is a certain size limit for the image processed by the recognition method; (2) Direct use of character information for license plate location and detection requires character detection on the entire image. In the scene with complex background or some characters in the background image, false detection is more likely to occur, which increases the computational complexity; (3) The parameter sharing in the deep learning network in the algorithm is not sufficient enough, that is, the different processes of the license plate recognition method are sufficient Shared network structure and network parameters

Images