Incorporating Attention Mechanism, Dense Connection Blocks, and Multi-Scale Reconstruction Networks for Open-Set Hyperspectral Image Classification

Hyperspectral image classification plays a crucial role in various remote sensing applications. However, existing methods often struggle with the challenge of unknown classes, leading to decreased accuracy and limited generalization. In this paper, we propose novel deep learning framework called IADMRN, which addresses issue class handling hyperspectral classification. IADMRN combines strengths dense connection blocks attention mechanisms extract discriminative features from data. Furthermore, it employs multi-scale deconvolution reconstruction sub-network enhance feature provide additional information for To handle utilizes an extreme value theory-based model calculate probability membership. Experimental results on three public datasets demonstrate that outperforms state-of-the-art terms both known classes. show proposed outperform several methods, outperformed DCFSL by 8.47%, 6.57%, 4.25%, MDL4OW 4.35%, 4.08%, 2.47% Salinas, University Pavia, Indian Pines datasets, respectively. The is computationally efficient showcases ability effectively classes tasks. Overall, offers promising solution accurate robust classification, making valuable tool