Design Of The Embedded License Plate Recognition System Based On Artificial Neural Network

In recent years,with the growth of the national economy,the number of cars more and more,the number of vehicles in some cities even reached a supersaturated state.The increase in the number of vehicles will inevitably bring many traffic problems,such as traffic congestion,parking spaces scarce,parking management confusion and so on.For these problems the traditional way of traffic management has been unable to meet the management requirements,then intelligent traffic management system came into being.Intelligent transportation system uses advanced information technology,computer technology,sensor technology and image processing technology to achieve the vehicle automation management.As an important part of intelligent transportation system,license plate recognition plays an important role in traffic management.Therefore,it is very important to study the license plate recognition technology.This paper designs a set of embedded license plate recognition system based on artificial neural network by understanding the latest developments of license plate intelligent identification technology,reviewing related papers and reading relevant books.The system uses OpenCV image processing database to simulate and realize the license plate recognition.This article mainly from the following aspects of the license plate recognition system design.Embedded platform design: including embedded hardware design and software design.The hardware section includes the CPU module,the camera module and the communication module.The software part includes GUI design,cross compilation,OpenCV configuration and Ubuntu system installation.License plate positioning: license plate positioning in the license plate recognition system plays a vital role.Based on the comparison and analysis of the existing positioning methods,the initial positioning of the license plate based on the combination of character edge detection and mathematical morphology is adopted,and then the SVM classifier is used to classify the license plate.Candidate license plate picture screened out the real license plate to complete the precise positioning of the license plate.Character Segmentation: Character Segmentation is for character recognition.There are many methods of character segmentation(such as vertical projection method,clustering method,template matching method and so on),but any kind of single character segmentation method has some aspects of the defect,so the use of template matching and vertical projection The method utilizes the projection feature of the character to search the right boundary of the second character in the projection map,and then uses the character geometry to divide the first and second characters.The last five characters using the template matching method,looking for the smallest error of the template to split the characters.Character recognition: Before entering the character recognition,it is necessary to extract the characters first.After analyzing the various features,this paper chooses the block statistical features and the projection features as well as the character characteristics.The extracted character features are sent into the artificial neural network for training,and the neural network model which can identify the license plate characters is obtained.