Hyperspectral Band Selection Using Moth-Flame Metaheuristic Optimization

Metaheuristic optimization algorithms have been popular band selection methods in hyperspectral imaging (HSI) over the years due to their ability to find good solutions in reasonable time frames. Nature-inspired metaheuristics imitate processes in nature to determine solutions and should be investigated in detail stochastically. In this research, the authors propose to use moth-flame optimization (MFO), inspired by the flight of moths around artificial light sources for HSI band selection. The MFO algorithm explores a search space with moths as search agents that circle inwards logarithmically to a flame or artificial light. At every step, the best solution is updated until the exploration is complete. In this paper, a pilot study of MFO based band selection for Indian Pines HSI benchmark dataset is presented. The results of MFO are compared with particle swarm optimization (PSO) and Genetic Algorithm (GA). Preliminary results show that MFO is a promising strategy for HSI band selection when compared to PSO and GA.