Optimum band selection in hyperspectral imagery using swarm intelligence optimization algorithms

Despite rich and fine spectral information of hyperspectral imagery, curse of dimensionality and Hughes phenomenon affect the land use/cover classification accuracy of such images. In this situation, optimal feature/band selection based on optimization procedures has high potential to improve the accuracy of hyperspectral image pattern recognition and classification. Among other optimization techniques, Meta heuristic optimization algorithms such as Swarm intelligence-based methods are so capable in solving feature/band selection problems. This paper evaluates the potential of Firefly algorithm (FA) and Particle Swarm Optimization (PSO) as representatives of swarm intelligence-based methodologies in optimal band selection and dimensionality reduction of hyperspectral imagery. Implementation results of Firefly algorithm and PSO in the case of AVIRIS hyperspectral image classification is compared with Genetic algorithm as another well known Meta heuristic optimization method. The preliminarily results confirm the high capabilities of Firefly algorithm and PSO for solving optimal feature/band subset selection.