Aiming at the problems such as high inter-class similarity,large intra-class diversi-ty,and large difference of remote sensing images at different scales,which lead to the limited accuracy of scene classification tasks,a multi-scale feature fusion based remote sensing image scene classifica-tion algorithm is proposed.Firstly,MobileNetV2,a lightweight network,is used as the backbone net-work to achieve higher classification accuracy with fewer network parameters.The network improves the input channel through one-dimensional convolution.The multi-scale feature fusion module de-signed in this paper can capture the high-level features and low-level features of remote sensing ima-ges,and the fusion of multi-scale features of remote sensing images can alleviate effectively the prob-lem of large differences in remote sensing images at different scales.The effectiveness of the proposed method is verified by experiments on three open remote sensing datasets.
关键词
遥感影像场景分类/多尺度特征融合/特征融合/深度学习
Key words
remote sensing image scene classification/multi-scale feature fusion/feature fusion/deep learning
维普
万方数据