Research Of Character Redognition In License Plate Recognition System

With the social and economic development, the number of cars increasing, resultingin a lot, such as traffic accidents, traffic congestion, environmental pollution and trafficproblems, intelligent transportation systems as a major traffic management style is therise in the conditions of this background. The license plate recognition system as a corepart of the Intelligent Transportation Systems, is a powerful weapon to achieveintelligent management of its research and development. It has become a popularmodern traffic to solve the increasingly serious road traffic management and can beused to transport flow monitoring, highway toll points, airports, car parks, residentialand other major traffic roadblocks, convenient and effective vehicle monitoring andmanagement. License plate recognition system is mainly involved in a number of digitalimage processing, pattern recognition, computer vision and artificial intelligencetechnology to analyze the collected image, to obtain useful information, so as to achievethe intelligent management.Firstly, were the key technologies of image pre-processing, positioning, segmentationand recognition of license plate recognition system description, on the basis of twolicense plate character recognition technology, and finally250collected plate images asan experimental validation, recognition rate of93.2%, with better recognition results.In this paper, the work mainly involves the following aspects:(1) the principle of the license plate recognition system for the current license platerecognition system algorithm described in the experimental results, while themain algorithm in VC++6.0and Matlab;(2) analyze the advantages and disadvantages of the traditional artificial neuralnetwork in the license plate recognition applications, the introduction of roughset and artificial fish swarm algorithm is proposed to optimize RS-RBF neuralnetwork, the license plate character recognition algorithm based on artificialfish. Using rough set theory to eliminate the advantages of redundant attributes,character feature reduction, the use of artificial fish swarm algorithm has torely on fewer empirical parameters, convergence threshold and the weights ofthe advantages of speed training RBF neural network, The simulationexperiments results show that the license plate character recognition algorithmbased on AFSA optimization RS-RBF neural network can quickly andaccurately to recognize license plate characters;(3) for the defacement of license plates or incomplete license plate, the traditionalneural network identification methods can not effectively identify license platecharacter recognition algorithm based on the similarity of characters, thecharacter of local information to match the principles of the improvementneural network and experimental results are given.(4) on the basis of theoretical studies, based on the Matlab platform developed license plate recognition system, and gives the experimental results of licenseplate recognition, finally250collected plate images as an experiment to verifythe recognition rate is93.2%.