This paper proposes a dual-input model for learning SAR-optical cross-modal features for land cover classification. The model integrates image-level and feature-level fusion methods to enhance the performance of land cover classification. The approach uses principal component analysis (PCA) to reduce dimensionality and improve feature-level fusion. A channel attention module is incorporated to highlight important features and suppress irrelevant information. The model is validated on various public datasets, demonstrating significant improvements in classification accuracy. The method effectively combines SAR and optical data, leveraging their complementary information to enhance semantic segmentation. The model is designed to be compatible with most encoding-decoding structures for feature classification tasks. The results show that the proposed method achieves superior performance compared to existing multimodal fusion approaches. The study highlights the importance of multimodal data fusion in land cover classification and suggests future research directions in this area.This paper proposes a dual-input model for learning SAR-optical cross-modal features for land cover classification. The model integrates image-level and feature-level fusion methods to enhance the performance of land cover classification. The approach uses principal component analysis (PCA) to reduce dimensionality and improve feature-level fusion. A channel attention module is incorporated to highlight important features and suppress irrelevant information. The model is validated on various public datasets, demonstrating significant improvements in classification accuracy. The method effectively combines SAR and optical data, leveraging their complementary information to enhance semantic segmentation. The model is designed to be compatible with most encoding-decoding structures for feature classification tasks. The results show that the proposed method achieves superior performance compared to existing multimodal fusion approaches. The study highlights the importance of multimodal data fusion in land cover classification and suggests future research directions in this area.