The License Plate Character Recognition Based On Particle Swarm Optimization Bp Network Research

As a core key technology of intelligent transportation system(ITS), licence plate recognition technology(LPR) plays an important practical value in intelligent traffic management and surveillance link.In recent years,the research of LPR system has received more and more attention from domestic and foreign scholars. How to improve real-time speed and identification accuracy becomes a hot and difficult spot.With regard to some issues in license plate recognition process,we did a deep and painstaking research on this questions.The content involved in this thesis summarized in the following respects.Firstly,The thesis introduced some relevant algorithm of digital image processing technology(DIPT),like gray change,edge detection,median filter and so on, completing the image denoising. Licence localization used Sobel edge detection operator、iteration threshold segmentation、correcting lean plate and character segmentation used algorithm based on prior knowledge and projection.Secondly,We used license plate recognition method based on BP neural network and feature extraction based on a combination of coarse grid and the neural network.Finally,this thesis used BP neural network training based on particle swarm optimization.This method combined particle’s global searching characteristics and BP neural network’s local search characteristics. Obviously, in experiment process,the results showed that network convergence speed increased.Study and experimental analysis show that algorithm adopted by this thesis can well achieve licence plate localization, segmentation and recognition.It also improves character recognition rate,reaching90percent.Furthermore,this algorithm can real-time realize license plate recognition.Meanwhile,through the computer simulation and experimental analysis,we conclude that this theory has certain feasibility, providing a theoretical basis for further research on license plate recognition for the future.