Design And Implementation Of License Plate Recongnition System Based On Deep Learning

License Plate Recongnition(LPR)is a problem that many research scholars have been discussing.Many license plate recognition systems are marked as more than 99% online,but in fact the license plate is actually used by the system in the application industry.It has always been difficult to propose a solution that it will be fully adapted to various situations.Many systems can achieve a recognition rate of more than 90%.It is very good,and most of these systems rely on high-definition cameras to transmit high-definition images,once the resolution of the image is achieved.If the license plate is degraded or the license plate is defaced,the recognition rate of the license plate will be greatly reduced.Therefore,it is still an unsolved problem that how to effectively improve the recognition accuracy.Many recognition systems use template matching methods in research methods.In recent years,with the development of artificial intelligence,deep learning has made great progress in image recognition,text recognition,face recognition,and speech recognition.It also provides a theoretical basis for introducing deep learning into license plate recognition.In addition,in many developed countries,there are relatively advanced license plate recognition systems,which play a huge role in intelligent transportation.However,because the specifications of China’s license plates are very different from those of foreign countries,it can not be adopted directly in China due to the actural conditions and current situation to be applied.Therefore,it is still a very significant topic to develope a license plate recognition system suitable for China.This paper is focused on the deep learning license plate recognition system,which is a method of deep learning in the recognition of license plates.However,the pre-processing of license plates still adopts the traditional segmentation extraction method,because the traditional algorithms for segmentation are more mature and reliable.The range that can be improved is small.The focus of this paper is the fusion of several algorithms.In the recognition of license plates,Convolutional-Neural-Network(CNN)is used to identify the accuracy and speed of recognition.The Matlab environment was used in the development,and the convolutional neural network was established.With the modular programming idea,the function modules were developed by using the function of sub-functions.Finally,a set of visualization software based on Matlab was developed,and the latter was considered.Expanding the requirements,with the help of Matlab interface can generate functions that can be called by other languages such as C language,which can be easily embedded into other systems.The part about mixed programming is mentioned in the text but not as the focus of this paper.Through research,it is found that the application of convolutional neural network in license plate recognition can be effectively improved the recognition rate of the license plate in various environments such as pollution,insufficient illumination,etc.This recognition rate is improved by means of a large training character set.The more character forms included in the character set,the higher the recognition rate,the more the license plate character recognition rate can reach 95% or more.In addition,for the trained convolutional neural network,including the license plate extraction and pre-processing recognition speed can also reach less than 1S.