A Study of PSO and its Variants for Fractional Abundance Estimation in Hyperspectral Data

Particle swarm optimization is an evolutionary stochastic global optimization method that has gained popularity in several applications. It has been applied successfully for engineering applications which includes robotics, image processing, remote sensing, and electrical power applications. This study introduces the application of Particle swarm optimization and several of its variant for abundance estimation operation in spectral unmixing of hyperspectral data processing. There are five PSO variants which have been used here for abundance estimation task. They run over different number of iterations in order to calculate the average abundance error. Comparisons of the different PSO approaches at different number of iterations are also presented. It is found that the PSO approach with dynamic inertia weight along with social and cognitive component i.e., PSO-2 has minimum error and has been performing well among the discussed approaches.