| In order to solve urban traffic management,intelligent traffic management system came into being.License plate recognition system is one of the key technologies of intelligent transportation system.This thesis focuses on improving the efficiency and accuracy of license plate recognition systems,starting from improving the quality of license plate images and improving the efficiency of recognition algorithms.The main work is as follows:(1)In order to improve the quality of license plate images,this thesis analyzes the traditional denoising algorithm,and finds that the traditional denoising algorithm has poor effect on noise filtering in the image,and it is easy to cause the loss of detailed information in the image after filtering the noise.Firstly,the traditional median filtering algorithm is improved,which can dynamically change the filter window size according to the pixel nature of the image,and use this improved algorithm to denoise the original vehicle image and the edge image extracted by the improved Canny edge detection operator,and finally combine the two filtered images to obtain the final improved algorithm results.Comparative experimental results show that the improved algorithm can maintain good denoising ability under a large amount of noise,and can still maintain image details after filtering out noise.(2)In this thesis,the traditional license plate character recognition methods are compared and studied,on the one hand,it is found that the character recognition method based on template matching has a low ability to recognize similar characters,resulting in a decrease in detection accuracy;on the other hand,it is found that the character recognition method based on convolutional neural network has low recognition time efficiency.In order to improve the efficiency and accuracy of license plate character recognition,a recognition algorithm based on similar characters combined with template matching and convolutional neural network is proposed,which first uses template matching to identify similar characters,and then uses convolutional neural network to perform secondary accurate recognition of similar characters.The comparative experimental results show that the license plate character recognition accuracy of the improved algorithm reaches 98.55%,and the recognition time is only 37.8 milliseconds,which ensures a high recognition rate and greatly shortens the recognition time compared with the neural network method.which has certain practical and theoretical significance. |
