Accurate building height estimation is essential to support urbanization monitoring, environmental impact analysis and sustainable urban planning. However, conducting large-scale building height estimation remains a significant challenge. While deep learning (DL) has proven effective for large-scale mapping tasks, there is a lack of advanced DL models specifically tailored for height estimation, particularly when using open-source Earth observation data. In this study, we propose T-SwinUNet, an advanced DL model for large-scale building height estimation leveraging Sentinel-1 SAR and Sentinel-2 multispectral time series. T-SwinUNet model contains a feature extractor with local/global feature comprehension capabilities, a temporal attention module to learn the correlation between constant and variable features of building objects over time and an efficient multitask decoder to predict building height at 10 m spatial resolution. The model is trained and evaluated on data from the Netherlands, Switzerland, Estonia, and Germany, and its generalizability is evaluated on an out-of-distribution (OOD) test set from ten additional cities from other European countries. Our study incorporates extensive model evaluations, ablation experiments, and comparisons with established models. T-SwinUNet predicts building height with a Root Mean Square Error (RMSE) of 1.89 m, outperforming state-of-the-art models at 10 m spatial resolution. Its strong generalization to the OOD test set (RMSE of 3.2 m) underscores its potential for low-cost building height estimation across Europe, with future scalability to other regions. Furthermore, the assessment at 100 m resolution reveals that T-SwinUNet (0.29 m RMSE, 0.75 $R^2$$R^2$) also outperformed the global building height product GHSL-Built-H R2023A product(0.56 m RMSE and 0.37 $R^2$$R^2$). Our implementation is available at: https://github.com/RituYadav92/Building-Height-Estimation.

中文翻译：

---

我们有多高？使用 Sentinel-1 SAR 和 Sentinel-2 MSI 时间序列在 10 m 处进行大规模建筑物高度估计


准确的建筑物高度估计对于支持城市化监测、环境影响分析和可持续城市规划至关重要。然而，进行大规模建筑物高度估计仍然是一项重大挑战。虽然深度学习 （DL） 已被证明对大规模测绘任务有效，但缺乏专门为高度估计量身定制的高级 DL 模型，尤其是在使用开源地球观测数据时。在这项研究中，我们提出了 T-SwinUNet，这是一种先进的 DL 模型，用于利用 Sentinel-1 SAR 和 Sentinel-2 多光谱时间序列进行大规模建筑高度估计。T-SwinUNet 模型包含一个具有局部/全局特征理解能力的特征提取器、一个用于学习建筑对象常数和可变特征随时间变化之间的相关性的时间注意力模块和一个用于在 10 m 空间分辨率下预测建筑物高度的高效多任务解码器。该模型使用来自荷兰、瑞士、爱沙尼亚和德国的数据进行训练和评估，并在来自其他欧洲国家/地区的另外 10 个城市的分布外 （OOD） 测试集上评估其泛化性。我们的研究包括广泛的模型评估、消融实验以及与已建立模型的比较。T-SwinUNet 以 1.89 m 的均方根误差 （RMSE） 预测建筑物高度，在 10 m 空间分辨率下优于最先进的模型。它对 OOD 测试集（RMSE 为 3.2 m）的强烈泛化强调了它在整个欧洲进行低成本建筑高度估计的潜力，并有望扩展到其他地区。此外，100 m 分辨率的评估显示 T-SwinUNet （0.29 m RMSE， 0.75 $R^2$ ） 的表现也优于全球建筑高度产品 GHSL-Built-H R2023A 产品（0.56 m RMSE 和 0.37 $R^2$ ）。我们的实施方法可在以下网址获得：https://github.com/RituYadav92/Building-Height-Estimation。