Experimental results of different methods on the CDD datasets.

SAASNets indicates our proposed change detection networks. Res2net50 indicates the initial backbone network. CPMA indicates channel position multi-head attention module. LAFEM indicates feature extraction module with special attention embedding. RASAM indicates the attention feature fusion module with residual strategy. Ld indicates the dice loss function.

SAASNets指代本研究提出的变化检测网络（change detection networks）。Res2Net50指代初始骨干网络（backbone network）。CPMA指代通道位置多头注意力模块（channel position multi-head attention module）。LAFEM指代带有特殊注意力嵌入的特征提取模块（feature extraction module with special attention embedding）。RASAM指代带有残差策略的注意力特征融合模块（attention feature fusion module with residual strategy）。Ld指代骰子损失函数（dice loss function）。