Estimation of Uncertainty Using Entropy on Noise Based Soft Classifiers

Abstract

In remote sensing noise is some kind of ambiguous data that occurs due to some inadequacy in the sensing, digitization or data recording process. This paper examines the effect of noise clustering algorithm of image classification. In remotely sensed data the easiest and usual assumption is that each pixel represents a homogeneous area on the ground. However in real world, it is found to be heterogeneous in nature. For this reason, it has been proposed that fuzziness should be accommodated in the classification procedure and preserves the extracted information. Classification of satellite images are complex process and accuracy of the output is dependent on classifier parameters. This paper examines the effect of various parameters like weighted exponent ‘m’ as well as resolution parameter ‘\(\partial \)’ for noise clustering (NC) classifier. The prime focus in this work is to select suitable parameters for classification of remotely sensed data which improves the accuracy of classification output to study the behaviour of associated learning parameters for optimization estimation using noise clustering classifier. A concept of “Noise Cluster” is introduced such that noisy data points may be assigned to the noise class. In this research work it has been tried to generate, a fraction outputs of noise clustering based classifier. The remote sensing data used has been from AWiFS, LISS-III and LISS-IV sensors of IRS-P6 satellite. This study proposes the entropy, as a special criterion for visualising and evaluating the uncertainty and it has been used as an absolute uncertainty indicator from output data. From the resultant aspect, while monitoring entropy of fraction images for different values, optimum weighting exponent ‘m’ and resolution parameter ‘\(\partial \)’ has been obtained for AWIFS, LIIS-III and LISS-IV images and that is ‘m’\(\,=\,\)2.9 and ‘\(\partial \)’\(\,=\,\) \(10^{6}\), providing highest degree of membership value with minimum entropy value as shown in Table 1.