Character recognition using parallel BP neural network

In the automated license plate recognition system, many reading errors are caused by inadequate character recognition method. This paper presents a novel character recognition method of license plate number based on parallel BP neural networks. This will enhance the accuracy of the recognition system that aims to read automatically the Chinese license plate. In the proposed methodology, the character is binarized and the noise is eliminated in the preprocessing stage, then the character feature is extracted by using skeleton and the character is normalized to size 8*16 pixels. Finally, the character feature is put into the parallel neural networks and the character is recognized. The proposed method in character recognition is effective, and promising results have been obtained in experiments on Chinese license plates.