Remote sensing image classification based on multimodal fusion and multimodel selection

Aiming at the limitations of single-modal remote sensing image classification, this paper proposes a remote sensing image classification method based on multimodal fusion. By fusing the complementary information of optical and microwave images, cross-modal complementary feature fusion technology is used to enhance feature expression. Experiments verified on the WHU-OPT-SAR and PIE-RGB-SAR datasets show that the improved method increases the overall accuracy by 2.45% and the mean intersection over union by 3.13%, with significantly improved classification accuracy for categories such as farmland, urban areas, and roads.