Unsupervised Image Classification by Radial Basis Function Neural Network (rbfnn)

This work extends the application of Radial Basis Function (RBF) neural network for the unsupervised classification of images. The radial basis function (RBF) network enables non-linear transformation followed by linear transformation to achieve a higher dimension in the hidden space. If classification is done in a high dimensional space, it is more likely to be linearly separable as compared to that in low dimensional space. This is directly related to the capacity of the network to approximate a smooth input-output mapping. A radial basis function like a spherical Gaussian, is a function that is symmetrical about a given mean or center point in a multidimensional space. Radial Basis function Neural Network (RBFNN) is a general regression technique, which is suitable for both function mapping and classification problems. Initially random centers were generated and then the final ones were calculated using the k-means clustering algorithm. The RBF Network has been implemented on IRS 1C LISS-3 image of Kanpur and adjoining regions, India.