<p>Band selection is a proficient dimensionality reduction method designed to determine an appropriate subset of spectral bands that minimises redundancy while maximising informational content. Evolutionary algorithms are considered an efficient tool for band selection due to their robust search capabilities. Nevertheless, the majority of current band selection methods based on evolutionary algorithms depend on single-search-strategy frameworks, which may inadequately discover optimal band subsets, readily converge to local optima. Many multi-objective optimization algorithm yield non-dominated solutions that are unevenly dispersed throughout the objective space. This study introduces a multi-strategy, multi-objective evolutionary algorithm for band selection, designated as MSMOEA-BS. An unsupervised multi-objective band selection model is developed, utilising information entropy to quantify the information capacity of bands and Jensen-Shannon Divergence to evaluate the correlations between selected bands. A variance-driven switching mechanism is subsequently devised to adeptly direct the population’s search process, accompanied by three separate search techniques to ensure appropriate selection for the population. Moreover, enhancements are implemented to the crowding distance to improve the diversity of the distribution of the ultimately selected individuals. Ultimately, to assess the efficacy of MSMOEA-BS, experiments were performed on three standard datasets utilising Support Vector Machine and K-Nearest Neighbours classifiers. Comprehensive experimental findings validate the feasibility and efficacy of the proposed MSMOEA-BS in hyperspectral images band selection. The code related to the MSMOEA is publicly available on GitHub: <a href="https://github.com/yanbohu11/MSMOEA">https://github.com/yanbohu11/MSMOEA</a></p>