Memetic Ant Colony Optimization for Band Selection of Hyperspectral Imagery Classification

This paper proposes a novel memetic ant colony optimization (MACO) algorithm for band selection on hyperspectral imagery classification. The method incorporates filter method based local search and ant colony optimization (ACO) based global search to take advantage of both. Particularly, the local search fine-tunes the paths explored by the ants by adding the relevant bands and eliminating irrelevant/redundant ones. A comparison study to the filters methods (including Gain Ratio, ReliefF, AP based method, and FCBF) and the counterpart wrapper ACO feature selection on four hyperspectral imagery datasets demonstrates that MACO is capable of attaining competitive or better classification accuracy with fewer selected bands. The empirical results suggest that MACO is effective and efficient in identifying relevant bands while eliminating irrelevant/redundant ones.