A license plate detection method based on depth learning

Abstract

The invention provides a license plate recognition method based on depth learning, which can self-adaptively obtain license plate characteristic information under different scenes, and simultaneouslyimproves the accuracy rate of license plate recognition. The invention adds the inherent color attribute of the license plate into the YOLO model to construct a license plate detection model with strong generalization ability and high detection and positioning accuracy, detects the passing picture taken by the input bayonet monitoring device, and obtains the area position of the license plate. Then, the inclined license plate area is corrected by Radon transform, and the license plate area is fine-tuned by color and edge. Finally, the fine-tuned license plate area is sent to CRNN network whichis constrained by license plate coding rules to recognize the license plate number.

Description

technical field [0001] The invention relates to the technical field of image recognition in intelligent traffic control, in particular to a deep learning-based license plate recognition method. Background technique [0002] Motor vehicle license plates are used as the unique identification symbols for motor vehicle management by national traffic management departments. The detection and identification of license plate numbers play an extremely important role in intelligent transportation systems; the existing license plate recognition methods are usually divided into three steps: the first The first step is to use the characteristics of the license plate to quickly locate the position of the license plate through clues such as color, edge, and texture, and obtain the candidate area of ​​the license plate target; the second step is to use SIFT ((Scale-invariant feature transform, scale-invariant feature transformation) , LBP (Local Binary Patterns, local binary pattern) and o...

AI Technical Summary

Problems solved by technology

[0003] In order to solve the problems of poor real-time performance and low accuracy in the real environment of existing license plate recognition methods, the present invention provides a license plate recognition method based on deep learning, which can adaptively obtain license plate feature information in different scenarios, At the same time, the accuracy of license plate recognition is improved

Images