Multi-Path Interactive Network for Aircraft Identification with Optical and SAR Images

Aircraft identification has been a research hotspot in remote-sensing fields. However, due to the presence of clouds satellite-borne optical imagery, it is difficult identify aircraft using single image. In this paper, Multi-path Interactive Network (MIN) proposed fuse Optical and Synthetic Aperture Radar (SAR) images for on cloudy days. First, features are extracted from SAR separately by convolution backbones ResNet-34. Second, piecewise residual fusion strategy reduce effect clouds. A plug-and-play Attention Sum-Max module (IASM), thus constructed interact with multi-modal images. Moreover, multi-path IASM designed mix backbones. Finally, fused sent neck head MIN regression classification. Extensive experiments carried out Fused Cloudy Detection (FCAD) dataset that constructed, results show efficiency identifying under different thicknesses.Compared single-source model, multi-source model improved more than 20%, method outperforms state-of-the-art approaches.