The Study And Implement Of The License Plate Recognition

Vehicle license plate recognition in computer vision and pattern recognition technology is the important application in the field of intelligent transportation, is to achieve one important aspect of traffic management. It is on the highway, residential security management and parking fees and other projects occupies a very important position. In this paper, by understanding other predecessors in the license plate recognition system on the related theories and technologies, researching related to license plate recognition algorithm, improved the algorithm for certain shortcomings and deficiencies.Character positioning is an indispensable component for vehicle license plate recognition system. In this paper, on the basis of previous studies, proposed a direct positioning method of license plate characters is based on watershed algorithm. The method made full use of the correlation between the license plate character, direct the positioning of license plate characters, abandoned the vehicle license plate location this step. Experiments show that the method in more complex environments can still accurate positioning of the license plate characters, and solve the problem of an image contains multiple license plates.Character recognition is the most important step in the license plate recognition system, the key is to identify the choice of classifier. This article first of making the character size normalization, extracting character feature vector with coarse grid method, finally putting the feature vector input to the trained neural network for character recognition. On this basis, the article study the neural network technology in the character recognition applications with the use of particle swarm optimization neural network weights of thought, propose a new neural network optimization method, optimization the network weights and structure at the same time, the simulation shows that this method is more effective, solves the problem which the number of the nodes in hidden layer of neural network in a certain extent.