Keep Your Friends Close, and the Right Neighbours Closer: Disaster-Conditioned Kernel-Regularized Graph Attention for Building Damage Classification

亲近你的朋友，更要亲近对的邻居：用于建筑物损坏分类的灾害条件核正则化图注意力机制

Abstract: Disaster damage is spatial: buildings rarely fail in isolation. Yet using spatial context for damage classification remains surprisingly unerexplored, and many pipelines still rely primarily on per-building appearance cues even when the dominant uncertainty is spatially structured. 摘要： 灾害造成的破坏具有空间性：建筑物很少会孤立地倒塌。然而，利用空间背景进行损坏分类的研究仍未得到充分探索。尽管主要的各种不确定性往往具有空间结构，但许多流程仍然主要依赖于单个建筑物的外观特征。

Complicating matters, the right neighbourhood is not the same across events. Floods, hurricanes, and wildfires can exhibit very different clustering behaviour, making spatial reasoning valuable but easy to misuse - naive context aggregation can improve visual coherence while oversmoothing boundaries or propagating structured errors. 更复杂的是，不同灾害事件中“合适的邻域”定义并不相同。洪水、飓风和野火可能表现出截然不同的聚类行为，这使得空间推理虽然有价值，却也容易被误用——简单的背景聚合虽然能提高视觉连贯性，但也可能导致边界过度平滑或传播结构性误差。

We study this tension on xBD (the dataset used in the xView2 challenge) in a controlled post-localization, classification-only setup: each building is represented by a pre/post combined (PPC) patch cropped from the provided polygons, and spatial context is modelled with GPS-derived building graphs. 我们在 xBD（xView2 挑战赛中使用的数据集）上，通过受控的“定位后、仅分类”设置研究了这一矛盾：每个建筑物由从给定多边形中裁剪出的灾前/灾后组合（PPC）补丁表示，并使用基于 GPS 的建筑物图来建模空间背景。

Our approach keeps local evidence “close” by preserving strong spatial relationships in disaster damage patterns, while bringing only the right neighbours “closer” through a disaster-type-conditioned graph model that injects a learnable multi-scale spatial kernel prior into attention, allowing the effective neighbourhood scale to adapt across disaster types rather than being learned as a single global smoothing rule. 我们的方法通过保留灾害破坏模式中强有力的空间关系来保持局部证据的“亲近”，同时通过一种灾害类型条件化的图模型，仅将“正确的邻居”拉得更近。该模型将可学习的多尺度空间核先验注入注意力机制，使有效的邻域尺度能够根据灾害类型进行自适应调整，而不是将其作为单一的全局平滑规则来学习。

To discourage coherence-by-smoothing, we add a residual de-correlation loss that penalizes positive Moran’s I in prediction residuals. 为了抑制“通过平滑实现连贯性”的倾向，我们增加了一个残差去相关损失函数，用于惩罚预测残差中存在的正向莫兰指数（Moran’s I）。

We evaluate the method under event and dataset shift with a leave-one-event-out (LOEO) protocol on xBD and cross-dataset transfer from xBD to Ida-BD. The model improves macro-F1 and substantially reduces residual spatial autocorrelation under zero-shot event shift, indicating better use of spatial context rather than naive smoothing and enabling more reliable transfer to unseen events within known disaster types. 我们在 xBD 上使用留一事件交叉验证（LOEO）协议，并在从 xBD 到 Ida-BD 的跨数据集迁移场景下评估了该方法。结果表明，该模型在零样本事件迁移下提高了宏观 F1 分数，并显著降低了残差空间自相关性，这证明了模型更好地利用了空间背景而非简单的平滑处理，从而实现了在已知灾害类型内向未见事件的更可靠迁移。