Attention-driven multi-feature fusion for hyperspectral image classification via multi-criteria optimization and multi-view convolutional neural networks

Sofiene Abidi, Akrem Sellami

Abstract: Highlights•Spectral feature selection method through multi-criteria optimization.•3D Convolutional Autoencoder effectively captures spatial information.•Multi-view fusion with 3D CNN improves spectro-spatial classification.•Attention mechanism dynamically weighs the importance of spectral and spatial features.•Superior classification performance demonstrated on real HSI datasets.

External IDs:dblp:journals/eaai/AbidiS24