Optimised Firefly Algorithm for Adaptive Band and Feature Selection in HyperSpectral Image Classification

This paper introduces an optimized approach for hyperspectral image classification through Adaptive Band and Feature Selection, utilizing an Improved Firefly Algorithm (IFA). By choosing an ideal selection of spectral bands, the suggested IFA successfully lowers the dimensionality and complexity of the classification task. based on objective functions that maximize divisions between class, such as the Jeffreys–Matusita (JM) distance, eliminating the need for an embedded classifier within the selection process. Additionally, the IFA is applied for feature selection, identifying the most discriminative features from a combination of filter-based feature sets. The method is tested on the Indian Pines hyperspectral image dataset, acquired by AVIRIS sensors. Experimental results demonstrate that the IFA significantly outperforms traditional methods, including the standard Firefly Algorithm, PSO, and PCA, as to classification accuracy and computational efficiency. A MSVM is used to evaluate the classification performance of the selected bands and features, further showcasing the potency of the proposed approach in hyperspectral image classification. This version integrates the results and comparisons with other methods while maintaining clarity and conciseness.