Hyperspectral Band Selection using SRL-SOA Algorithm

Hyperspectral image analysis is an important area in remote sensing, capturing detailed spectral information, thereby providing the possibility of distinct identification and categorization of the material. However, the vast data volume of HSIs leads to enormous memory and processing constraints. Our proposed methodology addresses these challenges using a three-fold approach: (i) band selection by an autoencoder-based approach,we first select the most informative spectral bands to decrease data redundancy. We leverage Optimized Self-Representation Learning with Sparse 1D-Operational Autoencoder (SRL-SOA) to achieve precise and efficient band selection for hyperspectral image analysis. This approach improves feature representation while minimizing redundancy, paving the way for improved downstream processing and reconstruction. The band selection results were evaluated by comparing with various band selection algorithms like PCA, ISSC etc. The experimental findings demonstrate the superior efficacy of the proposed model in comparison to other models.