Multimodal data fusion can provide valuable and diverse information for remote sensing image segmentation. However, different modal data have different feature distributions, which causes some conflicts and redundancies in cross-modal feature fusion. In addition, existing multimodal fusion networks usually adopt a two-branch structure with a large number of parameters and high computational cost. To address these problems, we propose a multiscale cross-modal knowledge transfer network (MCKTNet) for remote sensing image segmentation. First, we use a cross-modal migration learning method that combines channel discrete loss and spatial discrete loss to facilitate cross-modal migration of geometric and semantic features and reduces redundancy by minimizing the differences in feature distribution. Then, we use the polarized cross-self-attention mechanism to establish long-range correlations between different modal features across spatial and channel dimensions, and only a small number of parameters need to be added to achieve complementary cross-modal feature fusion. Finally, to accurately capture object edges, we propose a multiscale edge perception module to optimize the edge details in the prediction results at the pixel level. Extensive experiments demonstrate that the proposed method is effective, robust, and generalizable, and achieves state-of-the-art performance in multiple remote sensing image semantic segmentation tasks with only 13.51 million parameters. The code will be available athttps://github.com/NUAALISILab/MCKTNet.

Shadow detection is a basic task of remote sensing image analysis, but it is often seriously disturbed by vegetation, water bodies, and black objects. It is observed that vegetation and dark objects often show a dark look in visible bands but brighter in the near-infrared (NIR), and is also noticed that the reflection of inland water bodies in the green band is stronger than that in the blue band. Taking advantage of these physical properties and combining them with the bluish and dark appearance of shadows, we propose a simple but effective shadow detection method for multispectral remote sensing images. These physical properties are used to create transformation models that suppress features such as vegetation, water bodies, etc., but at the same time enhance shadows. Then, we transform the shadow representation into a color space to generate candidate shadows using dominant color components. To separate shadows from the others, we propose two indexes, the normalized Color Difference Composite Index (CDCI) and Color Purity Index (CPI), and fuse them to achieve shadows and their confidence. The experimental results indicate that the proposed method can effectively detect the shadows in multispectral images and outperforms the state-of-the-art approaches.