CSDACD: Domain-adaptive Change Detection Network for Cross-seasonal Remote Sensing Images

Change detection (CD) is a crucial task in Earth remote sensing, and it plays a significant role in analyzing changes in a region. However, bitemporal images acquisition may cause domain shift problems due to the seasonal difference. Most cross-seasonal CD methods currently in use make it difficult to capture the impact of domain shift. In contrast, domain adaptation (DA) methods struggle with fuzzy semantic correspondence and suboptimal use of generated data, which leads to incomplete target images. To address these problems, we propose CSDACD, an end-to-end DA CD network that integrates DA images into the CD network. In addition, we introduce the semantic change alignment (SCA) module to include the registration and semantic change information of bitemporal images and design three data fusion modules to enhance the adaptability of DA and CD networks. Our experiments show that the CSDACD outperforms the state-of-the-art (SOTA) methods in cross-seasonal CD tasks, achieving an F1 score of 98.73% and an IoU of 97.49% in the CDD dataset. Source code and models are available at https://github.com/24kironhead/CSDACD.