| 摘要 |
PURPOSE:To obtain the method and device to identify with constancy for the dislocation, rotation, scale conversion and gradation of money. CONSTITUTION:The weight coefficient of constancy is stored beforehand concerning plural pieces of money by using a neural network. The input pattern concerning the money to be identified is obtained and the money is identified at the neural network by using the stored weight coefficient. Since the weight coefficient is learned at the neural network while there is a instructor beforehand, the money to be identified can be identified. |
