Research On License Plate Recognition Technology Based On Neural Network

The Automatic License Plate Recognition System(LPR)utilizes digital image processing techniques to locate and identify the characters on the license number and output the result as a text string or other type of data format that can be easily understood by the operator.The LPR system has been applied to a variety of surveil ance applications,such as automated electronic toll collection(ETC),automatic parking attendants,which require the automatic control of the presence and identification of motor vehicles by their license plate numbers.Because LPR application brings great convenience to modern management of highway and urban traffic,and further promote the development of pattern recognition field,neural network is the most noticeable direction of pattern recognition in recent years,especially in depth learning Neural network,in which convolution neural network(CNN)is a common depth learning architecture,which has been applied in the field of speech analysis and image recognition in recent years,and has been applied.It has two advantages: weight sharing and sparse connection,so that the network structure and biological neural network closer to the weight sharing reduces the complexity of the network model,and a significant reduction in the number of weights.In the first chapter of this paper,the author discusses the main application technologies of automobile license plate recognition and the development trend of domestic and foreign.The second chapter introduces the relevant technology of license plate image processing,including image grayscale,image enhancement,binarization,edge detection and other pre-processing methods;and then for the most important license plate location,this paper introduces several commonly used Including the texture feature analysis method,mathematical morphology method,wavelet transform and so on.Then we introduce the character segmentation algorithm based on the projection feature and the connected domain feature.At last,we introduce the characters recognition technology of the license plate,The third chapter introduces the development,structure,derivation and implementation of convolution neural network in detail,and the training process.Chapter 4 deals with the implementation of license plate recognition in a step-by-step way.The fifth chapter introduces the experimental results and the analysis of the results of the license plate recognition system.The sixth chapter summarizes the content of the paper and looks forward to the future development of the license plate recognition.This paper is based on the deep learning neural network to design the license plate recognition system.Compared with the traditional shal ow neural network,the depth learning model has stronger learning ability and can improve the recognition rate of the license plate characters.In this paper,a series of image preprocessing techniques are used to extract the license plate information from the license plate.Because the single positioning method is difficult to locate accurately under complex conditions,this paper adopts the combination of the texture and color characteristics of the license plate.The location of the license plate after positioning through the pre-processing will inevitably contain some noise points,taking into account this point in the design of a combination of prior knowledge of the vertical projection method of segmentation on the positioning of the positioning of the license plate,And then uses the method of convolutional neural network to recognize the segmented characters and obtain the recognition result.The structure of CNN is based on the classical CNN structure,which increases the number of the feature map and the number of neurons in the output layer by combining the characters of the license plates and the difficulty of Chinese character recognition.From the experimental results we can see that the combination of a variety of pretreatment methods can be more accurate positioning,segmentation license plate,in addition,the use of convolutional neural network method to identify characters,compared to traditional neural network character recognition rate 5%.