Research And Implementation Of License Plate Recognition Algorithm Based On Deep Learning

With the rapid development of technologies such as the Internet of Things and artificial intelligence,smart cities are rapidly advancing with these technologies.As an important component of smart city,Smart Transportation has achieved the sharing of traffic information,real-time monitoring of road traffic conditions and comprehensive management thanks to the rapid development of information technology.Among them,license plate recognition plays an important role in traffic safety management and urban service systems.This paper combines deep learning technology with license plate as the research object to study the methods of license plate detection and license plate character recognition.This paper summarizes the common algorithms and improves the accuracy of license plate detection and license plate character recognition based on the corresponding algorithm.The main work is as follows:In order to solve the problem that the license plate image is not highly accurate due to the influence of complex environment such as light,weather,and shooting angle,This paper studies an improved method for license plate inspection based on regression target detection algorithm.This method selects deep residual network as the front network,adds multi-scale fusion feature layer,the license plate data set input into the network training model and the model is used to detect the license plate area.After obtaining the license plate area,in order to improve the accuracy of license plate detection,the license plate fine positioning method is added.Extend the license plate area and perform adaptive binarization on the area,analyze the connected area of the image to find the character frame,and use the random sample consensus algorithm and Sobel operator to determine the boundaries of the license plate.In the traditional license plate recognition algorithm,there are problems that the recognition rate of Chinese characters with complex shape structure is not high and the license plate characters are difficult to segment.This paper studies a convolutional neural network structure with continuous convolution and multi-label classification.This structure can extract more image features and achieve end-to-end character recognition.A large number of license plate images are generated,and the image is processed by adding noise and affine transformation to simulate the license plate images obtained in complex scenes.The data set is input into the network for training to obtain a license plate character recognition model.In order to further improve the accuracy of license plate recognition,this paper use the skew correction algorithm based on the direction field to preprocess the license plate image to be recognized,and then use the network model to identify the license plate characters.The license plate detection network and character recognition network model are applied in practical applications.This paper designs and implements a license plate recognition system based on deep learning.The system loads the trained model to realize the functions of license plate detection and character recognition.Experiments are performed in the data set to verify the good performance of the system in detection and character recognition.