Abstract: Remote sensing image scene classification plays a crucial role in providing real-time surface information, particularly for complex environmental situational awareness. However, the discrepancy between taining and testing data in remote sensing images often leads to poor generalization performance of the models. To address the issue of differences between the target domain and the sowrce domain in scene classification tasks, this paper proposes a feature contrast domain adaptation network, which facilitates remote sensing image cross-domain scene classification. This network can extract the domain-specific style features for remote sensing images and construct the domain contrast features transformation method to achieve cross-domain scene classification of remote sensing images. Experimental evaluations conducted on publicly available remote sensing datasets demonstrate that the proposed method outperforms the other domain adaptation methods.