Multimodal Electromagnetic Vortex SAR Image Fusion using CNN

Due to the hollow characteristics of vortex electromagnetic beams, target information may be missing to varying degrees in electromagnetic vortex SAR images under different modes. The target information can be restored using image fusion methods. Traditional pixel-level fusion methods require precise setting of weights, which may result in loss of information from the source image. This paper proposes a Multimodal Electromagnetic Vortex SAR Image Fusion using CNN. First, training is conducted using a large-scale RGB-D dataset (NYU-D2), and then fusion tests are performed using electromagnetic vortex SAR images of different modalities. The comparative experimental results show that this method can integrate electromagnetic vortex SAR images of different modes and recover target information, thereby producing high-quality SAR images.