Abstract:

The Hopfield neural network is one of the commonly applied neural networks in the artificial intelligence fields. In this paper, the pattern recognition of selected traffic signs is presented using the  discrete Hopfield neural network. The correlation is explored between the pattern recognition success rate and the level of noise addition and rotation as corruption mixed with the given patterns. Some new concepts, such as the complexity of traffic signs and recognition rate, are defined and employed in this work. The analytical and test results well indicate the good potentials of the Hopfield neural network in the identification of traffic signs and other similar patterns.

Key words: Hopfield neural network；traffic signs；image complexity;antinoise performance；pattern recognition rate