Research For Relative Techniques Of License-Plate Recognition System

The automatic license-plate recognition(LPR) system is one of the most important parts of ITS (Intelligent Transport System), and it is also an attractive subject.Generally, LPR system mainly comprises modules of plate location, character isolation and character recognition. The article does a lot of research on LPR system, brings forward some effective strategies and completes some experiments. Based on these a LPR software is developed.In the dissertation, it discusses some important image processing methods in LPR, then does some deep research on license-plate location, hopes to find some location methods that have high location accuracy and great computing speed by the research. In the part of character isolation, it uses Hough transformation and image rotation as a slant rectification method, then it researches Otsu image binarization methods and adopts a vertical segment method grounds on projection and transcendental knowledge.Template matching and BP neural network are used to distinguish the characters. Firstly, some methods are proposed to improve the convergent speed and precision of BP neural network. Secondly, the advantages and disadvantages of template matching and neural network are compared and analyzed, then these two methods are combined and some classifiers using a multilevel-multiclassification scheme are designed. The result shows that it is more effectively in character recognition.Finally, this article introduces the software architecture and performance of the license-plate recognition system. It also put forward some improvement for the future research.