Designing fuzzy rule based classifier using self-organizing feature map for analysis of multispectral satellite images

TL;DR

This paper introduces a novel fuzzy rule-based classifier for multispectral satellite images, utilizing self-organizing feature maps to generate prototypes and context-sensitive reasoning for improved classification accuracy.

Contribution

It presents a new method combining SOFM and fuzzy rules with context-sensitive inference, enhancing classification performance on satellite imagery.

Findings

Performance surpasses existing methods on multiple datasets

Context-sensitive reasoning improves classification accuracy

Fuzzy rule generation effectively captures feature space regions

Abstract

We propose a novel scheme for designing fuzzy rule based classifier. An SOFM based method is used for generating a set of prototypes which is used to generate a set of fuzzy rules. Each rule represents a region in the feature space that we call the context of the rule. The rules are tuned with respect to their context. We justified that the reasoning scheme may be different in different context leading to context sensitive inferencing. To realize context sensitive inferencing we used a softmin operator with a tunable parameter. The proposed scheme is tested on several multispectral satellite image data sets and the performance is found to be much better than the results reported in the literature.