MultiGO: An unsupervised approach based on multi-objective growth optimizer for hyperspectral image band selection

Abstract

Hyperspectral imaging (HSI) plays a crucial role in extracting discriminative spectral-spatial features for accurate land cover classification. However, HSI datasets often suffer from the presence of irrelevant and redundant spectral bands, leading to the Hughes phenomenon and increased computational complexity. To address this challenge, this paper proposes an unsupervised approach based on the multi-objective growth optimizer for hyperspectral image dimensionality reduction. The proposed method leverages the learning phase and reflection phase of the growth optimizer to balance exploration and exploitation strategies. By incorporating information richness, reducing redundancy, and considering spatial features, the growth optimizer selects the most informative and significant spectral bands. The approach simultaneously optimizes three objective functions using the growth optimizer, creating trade-offs among them. Extensive results demonstrate the effectiveness and superiority of the proposed method in achieving dimensionality reduction and preserving the essential information in hyperspectral images. Ultimately, four machine learning classifiers, namely support vector machine, random forest, K-Nearest Neighbors, and decision tree, are applied at the pixel level for hyperspectral image classification. Moreover, the proposed method shows a significant improvement compared with five state-of-the-art techniques (bat algorithm, archimedes optimization algorithm, particle swarm optimization, harmony search, and genetic algorithm), with overall accuracy equal to 80.95 %, 92.63 %, and 90.30 % on three benchmark hyperspectral datasets namely Indian Pines, Pavia University, and Botswana, respectively.