Particle Swarm Optimization (PSO) based approach for Classification of Remote Sensing Images

Dimensionality reduction is a major task in remote sensing images. Feature selection is applied for performing dimensionality reduction. It selects the spectral features(i.e. Bands) and find a feature subset that preserves the semantics of the hyperspectral image. Based on particle swarm optimization (PSO), this paper proposes multi-objective functions for selecting the spectral feature subsets for classification. The multi-objective function select feature subsets based on Jeffries Matusita(JM) distance and classifier(i.e. SVM). This paper performs optimal band selection and dimensionality reduction of hyperspectral imagery. The goal of the system is to perform spectral feature selection using particle swarm optimization (PSO) based multi-objective function. The system implements multi-objective functions which performs spectral feature selection (i.e. most informative bands) from the hyperspectral image dataset. These selected features are further used for evaluating the overall classification accuracy.