Chinese License Plate Recognition Based On Convolutional Neural Network

Accompanying the rapid development of China's social,economy and the improvement of residents' living standards,more and more families choose to buy cars as a travel tool,and then there cames vehicle management issues,such as vehicles driving on the road,entering and leaving residential areas,parking in parking lots,etc.,need to be supervised,and the easiest way to identify vehicles is to identify the license plate number.Due to the large number of vehicles and strong fluidity,only manual identification is difficult,accompanied by the development of artificial intelligence and hardware equipment.The improvement of the technology has resulted in the appearance of license plate number recognition technology based on machine learning and deep learning.The license plate recognition algorithm proposed in this paper is divided into three parts: the license plate positioning part,the license plate segmentation part,and the license plate recognition part.The license plate recognition is the most important part,and its accuracy is directly related to the performance of the entire system.In this paper,the license plate localization part uses an algorithm based on edge information and color information.The license plate character segmentation part uses a vertical projection method.In the license plate recognition part,traditional recognition methods include template matching and machine learning recognition algorithms.The template matching method is to match the segmented characters with the sample template and select the matching result with the highest degree of matching.Although the template matching method is simple and fast in recognition,it has a poor effect in recognizing blurred license plate images.Machine learning recognition algorithms generally use SVM,KNN,ANN and other common algorithms for recognition,and their recognition accuracy is higher than template matching algorithms.However,because Chinese characters are more complex than letters and numbers,under extreme conditions such as character defects,blurring and deformation the recognition effect is poor.Aiming at the problem of poor recognition of traditional Chinese license plates by traditional recognition methods,this paper proposes a new license plate character recognition algorithm.Compared with traditional algorithms,the license plate character recognition algorithm based on convolutional neural networks has a higher character recognition rate and faster recognition speed and wider application range in complex environments.