Neuro-Fuzzy Classifier for Astronomical Images

Abstract

In this paper, we sought to determine whether fractal parameters alone are good enough in classifying astronomical images. A fuzzy membership function which follows the model of a parabola was chosen for the purpose and the success rate was found to be 73.45%. Also, we have investigated how the grade of membership functions affect the performance of a neural network classifier. For this we included the parameter, spectral flatness measure in addition to fractal dimension and the grades of both the parameters were given as input features to the neural net. It could be observed that when grades were given as inputs to the classifier, performance of the classifier has increased to 80.53%.