This paper considers rotation-invariant pattern recognition systems using artificial neural networks. First, five neural pattern recognition systems are described. Next, they are applied to a variably rotated coin recognition problem to show those effectiveness. A 500 Japanese yen coin and a 500 Korean won coin classified in this paper have the same shape, size, and thickness, and have a similar pattern. In relation to such a fact, there was the report on misclassification between those coins in a newspaper.
From the results of computer simulation for coin recognition, considerations on computational complexity in hardware implementation of those systems and on modelling of our brains are described.
抄録全体を表示