Abstract: The Chinese suit tiles recognition system based on BPNN technology is designed to realize the automatically deal Chinese suit tiles machine to overcome a sort of drawbacks brought by artificial shuffle. In order to improve the accuracy of the recognition system, the training images are preprocessed: on the one hand, many kinds of standard features matrix are derived by the closing operation, dilation operation and erosion operation to increase the number of training samples, on the other hand, some regularities are found out and the color eigenvalues and the case eigenvalues are extracted, then the images and the BPNN are classified by these eigenvalues. The simulation results shows this technique is of high rate of classified accuracy and can meet the needs in the competition practical application of Chinese suit tiles.

Key words: BP neural network, image processing, classification, recognition, automatically deal Chinese suit tiles machine