Pattern Recognition for Partial Sensed Pictures by Plural Neural Networks linked as a Cascade and Application of Paper Currency

Abstract

Up to now, our group have proposed a new type of recognition method using neural network (NN) which is endowed with generalization and flexibility. Especially, this method is constructed by mask processing, mask optimization by genetic algorithm, and neuro-recognition board. In this paper, we discuss introduction of the NN to partial sensed pictures of the paper currency. However, the partial sensed pictures are different each other even though they are same category of the pattern recognition. When we recognize those pictures using a conventional single NN, it is supposed that the recognition ability is not enough in the market. Then we propose a new recognition system which is constructed by plural NNs linked as a cascade for the partial sensed pictures. First, we show a problem of recognition using the conventional single NN by the simulation results. Then, in order to solve the problem, we show a construction of the proposed recognition system. Finally, we discuss ability of the proposed system by the simulation results.