【摘要】 基于遥感影像进行洪涝灾害检测是获取灾害时空分布信息的有效手段之一。但大部分方法采用用户定义或选择的浅层特征作为检测依据泛化性较差。笔者提出了一种结合分类和语义分割的遥感影像洪涝灾害检测方法。将遥感影像输入到HRNet网络中进行分类，判断影像是否受灾；基于输入影像和分类检测结果，设计的语义分割方法对洪涝灾害实现更深层次的检测。实验证明提出方法的有效性，可为灾害救援及重建提供及时和可靠的信息。更多还原

【Abstract】 Flood disaster detection based on remote sensing images is one of the effective means to obtain information on the spatial and temporal distribution of disasters. However, most of current methods adopted user-defined or selected shallow features as the basis for detection, which often have poor generalization. A remote sensing image flooding disaster detection method combining classification and semantic segmentation was proposed. Remote sensing images are input to HRNet network for classification to determine whether the images are affected or not. The semantic segmentation method designed based on the input image and the classification detection result achieves deeper detection of flooding hazards. Experiments show that the proposed is effective, which can provide timely and reliable information for disaster rescue and reconstruction.更多还原