Classification using RBFNs based on fuzzy logic

There are four main parameters that can influence the quality of the multispectral image classification using the radial basis neural networks, the RBFNs receptive field, the number of the neural in the hidden layer, the center of each neural and the weights of the hidden layer to the output layer. The authors analysis and then give the details of how these four parameters affect the classification. To do this, the fuzzy logical theory is used. They give a graphic solution, instead of the functional analysis, which is used in most of the traditional methods. It is revealed that the RBFNs parameters optimization for the remote sensing image classification can be got without fail in the way of searching the receptive field inside its proper range. The reason is that the quality of the classification is consecutive to the change of each of these four parameters. Finally, a set of results is given to prove that the scheme is available and the conclusion is true