Recognition system of US dollars using a neural network with random masks
Fumiaki Takeda, Sigeru Omatu, Saizo Onami · 2005
We deal with bill money recognition and propose a downsizing method for neural networks by slab-like architecture. By this architecture, the same slab values which are the sum of input pixels may be obtained even if the inputs are different. To avoid this problem, we introduce random masks. We show that the recognition of US dollar bills which are much similar to each other and have the similar tone of colour can be done based on the neural network using a conventional bill money recognition machine and a 32-bit computer.