Hyperspectral Open-Set Classification via Frequency-Domain Rotation Enhancement and Multibranch Adversarial Routing

Abstract

Hyperspectral images have become indispensable for advanced material characterization and environmental monitoring, yet conventional analytical frameworks struggle with the evolving nature of spectral signatures in open-world scenarios. Open-set classification addresses this fundamental limitation by enabling recognition of both known and novel spectral categories during inference. Key technical barriers include rotational invariance in multiangle acquisitions, multiscale feature compatibility across spectral resolutions, frequency-domain discriminative decay, and interference from morphologically similar compounds. To overcome these challenges, we propose a frequency-domain multibranch adversarial routing open-set network integrating four core innovations: fractional Fourier transform layers for rotation-equivariant spectral localization, multibranch dynamic gate routing for uncertainty quantified hierarchical feature fusion, dual-frequency enhancement modules separating diagnostic spectral components through learned frequency gates, and a multiscale adaptive dynamic adversarial spectral mechanism enabling joint spectral–spatial attention refinement. The mathematical codesign of adaptive spectral operators and uncertainty-aware architectures establishes new theoretical foundations for robust open-set analysis in dynamic spectral environments.