HyperNet: Self-Supervised Hyperspectral Spatial-Spectral Feature Understanding Network for Hyperspectral Change Detection

TL;DR

HyperNet is a novel self-supervised hyperspectral change detection network that learns pixel-wise spatial-spectral features using a spatial-spectral attention module and a focal cosine loss, outperforming existing methods.

Contribution

The paper introduces HyperNet, a pixel-level self-supervised network with a spatial-spectral attention module and a focal cosine loss for improved hyperspectral change detection.

Findings

HyperNet outperforms state-of-the-art algorithms on six hyperspectral datasets.

The spatial-spectral attention module effectively captures spectral and spatial correlations.

The focal cosine loss improves training by emphasizing hard positive samples.

Abstract

The fast development of self-supervised learning lowers the bar learning feature representation from massive unlabeled data and has triggered a series of research on change detection of remote sensing images. Challenges in adapting self-supervised learning from natural images classification to remote sensing images change detection arise from difference between the two tasks. The learned patch-level feature representations are not satisfying for the pixel-level precise change detection. In this paper, we proposed a novel pixel-level self-supervised hyperspectral spatial-spectral understanding network (HyperNet) to accomplish pixel-wise feature representation for effective hyperspectral change detection. Concretely, not patches but the whole images are fed into the network and the multi-temporal spatial-spectral features are compared pixel by pixel. Instead of processing the two-dimensional imaging space and spectral response dimension in hybrid style, a powerful spatial-spectral attention module is put forward to explore the spatial correlation and discriminative spectral features of multi-temporal hyperspectral images (HSIs), separately. Only the positive samples at the same location of bi-temporal HSIs are created and forced to be aligned, aiming at learning the spectral difference-invariant features. Moreover, a new similarity loss function named focal cosine is proposed to solve the problem of imbalanced easy and hard positive samples comparison, where the weights of those hard samples are enlarged and highlighted to promote the network training. Six hyperspectral datasets have been adopted to test the validity and generalization of proposed HyperNet. The extensive experiments demonstrate the superiority of HyperNet over the state-of-the-art algorithms on downstream hyperspectral change detection tasks.