Research Of License Plate Recognition System Based On Fuzzy Hopfield Neural Network

Nowadays,with the enhancement of computer computing ability,the once-depressed neural network is once again reflected in people's vision.With the rapid development of neural network,it has made great progress in the field of artificial intelligence research.Nowadays,more and more scholars are devoting themselves to the research in this field,among which fuzzy theory and neural network are the focus of research.The license plate recognition technology studied in this paper plays a very important role in modern license plate management.How to combine artificial intelligence with license plate recognition and apply it to practical work is the main research topic of this paper.Hopfield neural network is a single-layer fully connected feedback neural network.This network introduces the concept of "energy function".In the process of network operation,attractors can be stored in the minimum energy point.Therefore,in the process of network energy decline,it will eventually become stable.It is based on the characteristics of Hopfield that we will use this network for license plate recognition.Because there are many uncertain factors in the process of license plate recognition,in order to avoid these uncertain factors which make the rate of license plate recognition too low,we will introduce the concept of fuzzy theory and combine the fuzzy theory with the neural network,which can not only solve the problem of poor learning ability of the fuzzy theory itself,but also improve the accuracy of recognition.In order to verify the feasibility of this algorithm,we will use another algorithm to compare with it,that is,the restricted Boltzmann machine.Restricted Boltzmann machine is a stochastic neural network,which can be used as feature extraction and classifier as a good recognition algorithm.In this paper,the combination of Hopfield neural network and fuzzy reasoning system is taken as the main body,and the algorithm is compared with the restricted Boltzmann machine.Finally,the following work and results are achieved: 1.Some methods of digital image processing and positioning are discussed.Appropriate methods are selected to process license plate image reasonably.The purpose is to make its features fully visible and easy to use the algorithm for recognition.2.The Hopfield network and Boltzmann machine are studied from the aspects of network structure,state,working mode,learning method and energy function.The applicability of the two networks in the direction of license plate recognition is determined.The parameters of Hopfield network and the number of neurons and threshold of Boltzmann machine used in license plate recognition are designed.3.This paper studies the fuzzy theory,determines the feasibility of adding fuzzy reasoning to the network,and combines the Hopfield network with the fuzzy theory,which improves the accuracy of license plate recognition to a certain extent.4.Visualization of fuzzy Hopfield network and restricted Boltzmann machine is realized by programming software.The advantages and disadvantages of two license plate recognition algorithms are analyzed and compared through actual system test.