License Plate Recognition Based Convolutional Neural Network

Convolutional neural network is a special neural network that can simulate the human brain function and can be applied to more areas.It is developing in recent years and compared with other methods,convolutional neural network has the advantages of global optimization,strong adaptability,complete theory and optimized scalability.Convolutional neural network is a hot research topic in the field of machine learning.However,as a new kind of technology,the applications of neural network in many fields still need to be explored and improved.In today's society,the intelligent transportation system using state-of-the-art technology to upgrade the traditional traffic system,has benefitted the society in the aspects of the efficiency and economics.With the development of communication technology and computer network technology,more and more countries pay attention to the vehicle license plate recognition system.The actual application of traditional license plate recognition methods in preprocessing will present shortcomings such as inaccurate license plate location and character segmentation errors.These shortcomings could affect the recognition effect and reduce the actual recognition rate.Besides,traditional image preprocessing of license plate recognition method is time-consuming.It is unable to cope with the real-time actual application and it is vulnerable to noise.Additionally,it is difficult to retain the original signal completely,which could further reduce the recognition rate.In this thesis,based on the analysis of the mechanism of the convolutional neural network method,convolution neural network is introduced into the intelligent transportation system.Its special structure with local weight sharing shows the unique superiority in image processing.Because of the feature that multi-dimension input image vector can be inputted directly,the effect of image recognition and processing can be better when apply this method,which could further avoid complexity problem of data reconstruction,feature extraction and classification process.Working with the license plate recognition based on convolutional neural network,this paper summarizes the research results of the domestic and foreign academic status about the convolutional neural network and introduces the basic concepts and basic principles of convolutional neural network.To improve the classical LeNet-5 neural network structure based on the analysis of former research and then applied the improved one into the license plate recognition problem.Then develop the software on MATLAB platform and finally complete the research of license plate recognition based on convolution neural network.Due to the superiority of neural network,the proposed license plate recognition method based on convolution neural network in this paper uses an improved version of LeNet-5 neural network structure.It optimizes the parameters of the convolution layer,the sampling layer of neural network and the recognition rate under particular environment.The overall recognition rate has been greatly improved,which has great social significance for the construction of intelligent transportation system.