Study Of Automatic Vehicle License-Plate Recognition Based On Wavelet Neural Networks

Wavelet neural networks can be regard as not only the function-linked networks based wavelet function, but also the extension of Radical basis function (RBF) networks. Compared with the normal feed-forward networks and RBF networks, it has different characteristic. So wavelet neural networks have great potential in the field of neural networks research. Wavelet neural networks combine the advantage of wavelet transform with the advantage of neural networks. It can be applied very well in the fields of function approximation, system identification, data compression, and so on.This paper applies wavelet neural networks in the system of vehicle license-plate automatic recognition and researches the effect of that application. We also apply back-propagate algorithm to recognize the same characters, and compare the effect of the two methods.As an important image analysis tool, Moment function has many comprehensive applications in the field of pattern recognition. The article uses the wavelet moment, the Zernike moment and the geometrical moment to extract the features of license-plate characters. We use the wavelet neural networks classify them, then compare the effects of all these moments in the extraction of character features.Finally, the paper introduces the software and hardware design in the steel number automatic recognition system.