SIFNet: A self-attention interaction fusion network for multisource satellite imagery template matching

Ming Liu, Gaoxiang Zhou, Lingfei Ma, Liangzhi Li, Qiong Mei

Abstract: Highlights•A novel Transformer-based deep fusion network (SIFNet) was proposed for remote sensing image matching.•Multiscale features were extracted by a pyramid network for each pixel.•The extracted features were fused by self-attention layers in Transformer for information interaction.•Two loss functions were developed to convert satellite imagery registration into classification and regression tasks for efficient matching purpose.