Design of two-stage cellular neural network filter for detecting particular moving objects
Katsuya Kondo, Hiromasa Morishita, Y. Konishi, H. Ishigaki · 2003
In this paper, we propose a new discrete-time cellular neural network (CNN) model for extracting a particular moving object, based on the CNN paradigm. Since the CNN-type filter has only spatially local interconnections and the number of connections between neurons is relatively few, the required computation in the learning phase is a reasonable amount. Instead, the output/input behavior of designed CNN filters is restrictive. Therefore it is significant that the structure of network model be discussed. The proposed CNN filter is formed by cascade connecting two 3-layer CNNs. In order to train the weighting factors, the backpropagation method is applied. Through simulations, it is shown that the target object is enhanced in the noisy environment.