Research Of License Plate Recognition Algorithm Based On Deep Learning

The intelligent transportation system not only increases the safety of the road,but also plays a huge role in the tracking of stolen vehicles,traffic monitoring,speed limit enforcement and automatic parking systems.The license plate recognition technology is an important part of the intelligent transportation system,and it is increasingly favored by researchers.The license plate recognition algorithm in this paper mainly includes two parts: license plate location and license plate character recognition.The first part: license plate location.The traditional license plate location algorithm has the problem of being unable to locate due to the influence of the environment such as illumination and camera shooting angle.The license plate location algorithm based on deep learning has the problem that the license plate area cannot be accurately located.In addition,existing license plate location algorithms may misidentify areas of the road that resemble license plates,such as signs on roads and billboards,as license plates.In order to solve the problems of the above license plate location algorithm,this paper proposes a license plate location algorithm based on the deep learning model YOLO-L.In this paper,the following two ways are used to more accurately locate the license plate area and improve the accuracy of license plate location:(1)K-means++ clustering algorithm is used to select the optimal number of license plates and vehicle area candidates,and then the results obtained by clustering are applied to YOLO-L model proposed in this paper.(2)Based on YOLOv2 model,YOLO-L model for license plate location is proposed.This model combines the characteristics of more different convolution layers to obtain better recognition results.At the same time,this paper uses the license plate pre-positioning algorithm to correctly distinguish the license plate area and the license plate-like area,and judge whether the identified license plate area is correct by discriminating whether the identified license plate area is within a certain identified vehicle area.The second part: license plate character recognition.The license plate character recognition method based on license plate character segmentation cannot correctly segment the license plate characters,which will directly affect the effect and efficiency of license plate character recognition.In addition,the natural environment of different lighting and license plate areas will also cause difficulties in segmentation and recognition of license plate characters.In order to solve the problems in the above license plate character recognition,after the license plate location,this paper proposes an end-to-end license plate character recognition method based on the deep learning network AlexNet-L.On the one hand,AlexNet-L network adopts the end-to-end method to reduce the influence of the license plate character segmentation result in the existing license plate character segmentation-based license plate recognition method on the license plate character recognition and improve the recognition efficiency of the license plate characters.On the other hand,end-to-end license plate character recognition based on AlexNet-L network can improve the recognition accuracy of license plate characters.In this paper,the proposed algorithm is comprehensively tested on two different data sets.Through the qualitative analysis and quantitative analysis of the license plate location algorithm,the experimental results show that the proposed license plate location algorithm can accurately locate the license plate area and is superior to other algorithms.The end-to-end license plate character recognition algorithm proposed in this paper can effectively improve the accuracy and efficiency of license plate character recognition.