Fuzzy rule-based hyperspectral band selection algorithm with ant colony optimization

The issue of band selection is extremely important in dealing with the plague of dimensionality in hyperspectral images. This study offers a hybrid band selection strategy based on the split-and-merge concept. This novel technique provides suitable band subgroups based on entropy and mutual information utilizing a fuzzy rule-based system without dismissing the real relevance of the band information. Then, using ant colony optimization, it finds the most promising hyperspectral bands from these subsets. On three prominent hyperspectral image data sets, four state-of-the-art techniques are compared with the suggested method to assess the importance of the proposed band selection strategy. In terms of kappa coefficient and overall accuracy, this approach outperforms others significantly.