Skip to content

FA3DHeat: A Frequency-Aware 3-D Heat Conduction Network for Hyperspectral Image Classification

2026 · IEEE Transactions on Geoscience and Remote Sensing · Vol 64, pp. 5527915-5527915 · 0 citations · 45 references

Abstract

Hyperspectral image (HSI) classification requires effective spectral–spatial representation learning, since spectral signatures and spatial structures are highly coupled. Although convolutional and attention-based methods have achieved promising performance, efficient modeling of spectral–spatial information in the intrinsic HSI cube remains challenging. Heat conduction-based modeling has recently emerged as a promising approach for efficient long-range information propagation in the frequency domain. However, existing heat conduction-based models are mainly designed for 2-D spatial features and cannot fully characterize the spectral–spatial structure of HSIs. To address this issue, we propose FA3DHeat, a frequency-aware 3-D heat conduction network for HSI classification. Specifically, a 3-D heat conduction operator (3DHCO) is first developed to extend the operation from the 2-D spatial domain to the 3-D spectral–spatial domain. To further exploit frequency-specific information, a multibranch frequency-aware (MBFA) module is designed to extract low-, mid-, and high-frequency representations with learnable soft masks (LSMs). Finally, an adaptive frequency branch fusion (AFBF) module is introduced to dynamically integrate these complementary branches with input-dependent weights. Extensive experiments on three public HSI classification datasets demonstrate that FA3DHeat achieves competitive performance compared with several state-of-the-art methods, validating the effectiveness of 3-D heat conduction for hyperspectral spectral–spatial representation learning. The code will be made publicly available at https://github.com/shidong66014/FA3DHeat

View source

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.