Research And Implementation Of License Plate Recognition Algorithm

In recent years, with the development of society and economy, our country motor vehicle to maintain a rapid growth trend, intelligent transportation system has become a frontier research topic in the field of transportation. The vehicle license plate automatic recognition technology (License Plate Recognition, LPR) automatic registration and verification can realize vehicle identity, is an important technology to realize the intelligent management of motor vehicle. Refers to thecomprehensive application of technology of digital image processing, computer vision, artificial intelligence and pattern recognition, license plate recognition system. General can be divided into image acquisition, license plate location,character segmentation, character recognition module. Among them, the license plate location and character recognition of license plate is a key technology in this system, in order to avoid premature and easy to fall into local optimum, this paper developed an efficient method of license plate location, improve the positioning accuracy; for more accurate license plate recognition, license platerecognition to improve the robustness, using the improved BP neural network algorithm for the license plate character recognition.This paper uses the gray image and global dynamic value of the two methods topreprocess the image of license plate, then according to the characteristics of the license area inherent, proposed by combining genetic algorithm and particle swarm optimization algorithm, the function is positioned correctly in the license plate region to the judgment condition assessment structure, selection operatorand genetic algorithm in the particle swarm algorithm, the improved algorithm of license plate location, license plate positioning accuracy. Treatment after the license plate on the positioning two value of the image, and then on the platemay tilt corrected, to detect the angle of inclination using projection method, then the image rotation correction by using coordinate transformation. Secondly,through the gray variation to determine the license plate boundary, through the vertical projection method to determine the license plate of the left and right boundary for character segmentation. Method and then use the prior knowledge and the vertical projection of the character segmentation is realized. Finally, ontwo of the most commonly used method of license plate characters recognition--template matching method and BP neural network method, based on improved BP neural network character recognition methods, the higher recognition rate of the algorithm is designed, and the experimental results of two methods are compared.Through Matlab simulation, the results show that the positioning algorithm,compared with the location and genetic algorithm and particle swarm optimizationresults, the improved algorithm can effectively avoid the interference of light,bumper, false license plate area, locate the license plate. And the400license plate images taken as experimental data, realize the license plate locationrecognition reaches98%, the improved BP neural network algorithm was alsosignificantly higher than the traditional template matching method and BP neuralnetwork method, the poor robustness of template matching method, the recognition rate of only91.33%, BP neural network is easy to fall into the local minimum, the recognition rate is95.01%, through the improvement of BP neural network algorithm, the recognition rate is up to97.37%, the license platecharacter recognition has been effectively improved, and achieved good results.