Multisource Remote Sensing Data Classification Based on A Dual Attention Fusion Network

Joint classification of multisource remote sensing data is a very meaningful yet challenging task. Especially when the actual scene composition is complex, spectral confusing categories overlapping and different appearances of the same object will bring great challenges for classification. In this paper, a novel Dual Attention Fusion Network (DAFNet) is proposed to solve the above problems. Firstly, a spectral attention block is designed to highlight or suppress the channel map, so as to better distinguish the spectral confusing categories. At the same time, we introduce a spatial attention block that integrates the features of all locations by weighted sum, to ignore the effect of spatial differences. Experimental results on real dataset show that the proposed method can effectively improve the classification results compared to other competitive works.