Optimum Band Selection of Hyperspectral Imagery Based on Particle Swarm Optimization

Nowadays, hyper-spectral remote sensing imaging systems are able to acquire several hundreds of spectral bands. Increasing spectral bands provide the more information for land cover and separate similarity classes, so classification accuracy potentially could increase. Nevertheless classification of hyperspectral imagery by conventional classifiers suffers from Hughes phenomenon. One of the solutions of this problem is using feature selection techniques. Traditional feature selection techniques have several limitations in performance and finding the global optimum subset of feature in this kind of remote sensing imagery. Recently optimization techniques based on meta-heuristic such as Swarm Intelligence (SI) and Genetic Algorithm (GA) are proposed. Particle Swarm Optimization (PSO) is one of the swarm intelligence techniques. PSO is inspired by social behavior of bird flocking or fish schooling. In this paper, a novel band selection algorithm based on Boolean Binary Particle Swarm Optimization (B-BPSO) is presented. The proposed method has advantage due to global search in too large space and don't trap in local optimum early. Evaluating obtained results from classification accuracy of AVIRIS image data set shows superior performance of proposed method as this obtains fewer numbers of bands and higher classification accuracy than evolutionary algorithms such as GA and older version BPSO.