Research And System Design Of License Plate Location And Recognition Based On Deep Learning

With the improvement of social and economic development level,the application scope of intelligent transportation system is more and more extensive.As an important part of intelligent transportation system,many researchers are constantly improving and optimizing the license plate recognition algorithm.The development of deep learning has pushed this research to a climax.Through the application of deep learning algorithm,the license plate recognition algorithm has been applied in a field.To a certain extent,the algorithm steps are simplified,and the accuracy is also improved.This paper aims at license plate recognition,aiming at the domestic open source CCPD data set,with the help of deep learning theory,designed a dual network model for the license plate location module,and then uses U-Net,FCN and SegNet to apply to the license plate location,and compares the above three models with the dual network model.Then,the CNN model is designed for the license plate recognition module.The CNN model solves the traditional pattern recognition characteristics of segmentation and recognition,and simplifies the implementation steps.Then,LeNet and ANN are applied to license plate recognition and compared with CNN model.The results show that the training accuracy of the dual network model and CNN model is as high as 98%.From the experimental results,the model designed in this paper is very effective for completing the goal of each module.The main contents and contributions of this paper are as follows.1.Research license plate location algorithm.Firstly,for the license plate location,firstly,the license plate image area is selected accurately by reducing the image size and data annotation,so as to reduce the redundant calculation of the car image,which reduces the training time of the model to a certain extent.Then,the structure layer and parameter definition of convolutional neural network model based on image segmentation are studied.A dual network model is designed to locate the license plate image to improve the accuracy of license plate location.Then,license plate correction algorithm is used to correct the license plate to improve the accuracy of license plate recognition.2.Research license plate recognition algorithm.According to the captured license plate image as the data set,the structure principle and parameter setting of convolution neural network are studied,and the convolution neural network model is designed for the accurate recognition of license plate characters,which reduces the training time of the model to a certain extent and improves the accuracy of license plate recognition.3.Simulation design of the system.This paper uses the deep learning training model framework combined with the Python UI interface to carry out the deep learning-based system simulation design,and compares the effect analysis with Baidu AI’s license plate recognition.Through the effect analysis,the system designed in this paper has a better effect on the number of recognition pictures.