HANet: A Hierarchical Attention Network for Change Detection With Bitemporal Very-High-Resolution Remote Sensing Images
This work addresses a specific bottleneck in remote sensing change detection, offering incremental improvements for applications like environmental monitoring.
The paper tackles the problem of class imbalance between changed and unchanged pixels in change detection with remote sensing images, proposing a progressive foreground-balanced sampling strategy and a hierarchical attention network (HANet) that achieves improved detection performance, as validated on two datasets with extremely unbalanced labels.
Benefiting from the developments in deep learning technology, deep-learning-based algorithms employing automatic feature extraction have achieved remarkable performance on the change detection (CD) task. However, the performance of existing deep-learning-based CD methods is hindered by the imbalance between changed and unchanged pixels. To tackle this problem, a progressive foreground-balanced sampling strategy on the basis of not adding change information is proposed in this article to help the model accurately learn the features of the changed pixels during the early training process and thereby improve detection performance.Furthermore, we design a discriminative Siamese network, hierarchical attention network (HANet), which can integrate multiscale features and refine detailed features. The main part of HANet is the HAN module, which is a lightweight and effective self-attention mechanism. Extensive experiments and ablation studies on two CDdatasets with extremely unbalanced labels validate the effectiveness and efficiency of the proposed method.