| The automatic license-plate recognition system is one of the most important parts of ITS (Intelligent Transport System), and it also is an attractive subject.This paper firstly introduces some other inspects of the license-plate recognition in brief, and then it pays much more attention to the license-plate characters recognition with multiple neural networks. The license-plate characters recognition with multiple neural networks include two main parts: the feature extraction of character and the construction of multiple neural networks.In the stage of characters' feature extraction, the paper fully uses the structural information and the statistical information. They all have some advantages in some inspects. The structural information of characters is sensitive to the framework and is good at distinguishing the similar characters. The statistical information of characters has the ability of robustness. By experimenting, several kinds of structural information and statistical information that are steady and can be supplemented for each other are chosen. After using all these character features as inputs, the paper constructs the multiple neural networks recognition system.This article uses the BP neural network to distinguish the characters. Firstly, it improves the neural network and the experiment datum proves that the improvement is good at the convergent speed and precision. Secondly, this paper bases the character recognition system on several neural networks that use different character features. The systems include the serial neural networks, the parallel neural networks based on judging logic and the parallel neural networks based on fuzzy operation. The experiment shows that the integration of multiple neural networks is better than single neural network in character recognition.Finally, this article introduces the hardware design for the license-plate character recognition. It also put forward some improvement for the future research.
|
PDF Full Text Request