A license plate location method and device based on deep learning

Abstract

The embodiment of the present application provides a method and device for locating a license plate based on deep learning. The method comprises: inputting the acquired image to be positioned into a feature extraction network, obtaining a feature map extracted by the feature extraction network, inputting the feature map into a marquee network, and obtaining a candidate area of ​​the license plate in the image to be positioned determined by the marquee network; The feature extraction network and the frame selection network are pre-trained through the sample license plate image; the candidate area is expanded, and the obtained expanded area is input into the classification network to obtain the classification result determined by the classification network; the classification network is used to obtain Vehicle texture features around the license plate area, determine whether the input extended area contains the classification result of the license plate area; when the classification result indicates that the extended area contains the license plate area, determine the license plate area of ​​the image to be located from the extended area. Applying the solutions provided by the embodiments of the present application can improve the accuracy of license plate location in complex scenarios.

Description

technical field [0001] The present application relates to the field of intelligent transportation technology, in particular to a deep learning-based license plate location method and device. Background technique [0002] With the maturity of digital image processing, pattern recognition and artificial intelligence technology, license plate recognition technology is also constantly improving. Among them, license plate recognition is the basis for realizing intelligent transportation system. Usually, the license plate recognition process includes three links: license plate location, character segmentation, and character recognition, and the license plate location is a very important link in the license plate recognition. [0003] Nowadays, there is no lack of high-accuracy license plate positioning technology, but most of them are based on the monitoring images of traffic checkpoints. In the existing positioning methods, based on edge detection, the license plate area can be...

AI Technical Summary

Problems solved by technology

However, in non-traffic checkpoint surveillance scenes similar to those shot by spherical cameras, the position, angle, size, and definition of the license plate area vary and the background is cluttered.

On the one hand, the pixel values ​​of the boundary of the license plate area and the pixel values ​​of the non-license plate area may be blurred and cannot be distinguished; on the other hand, there may be pixel areas similar to the boundary of the license plate area in the cluttered background, which may cause misrecognition, such as Traffic signs may be positioned as license plate areas

Therefore, the previous license plate positioning technology based on traffic checkpoints is difficult to achieve satisfactory accuracy on the images in the above complex scenes, and the accuracy of license plate positioning is not high

Images