Rauch Lukas, Braml Thomas
Institute of Structural Engineering, University of the Bundeswehr Munich, Munich, Germany.
Sci Data. 2025 Aug 25;12(1):1478. doi: 10.1038/s41597-025-05827-7.
We introduce Rohbau3D, a novel dataset of 3D point clouds that realistically represents indoor construction environments. The dataset comprises 504 high-resolution LiDAR scans captured with a terrestrial laser scanner across 14 distinct construction sites from high-rise buildings, all in various stages of shell construction or under renovation. Each point cloud is enriched with the scalar laser reflectance intensity, RGB color values, and reconstructed surface normal vectors. In addition to the 3D data, the dataset includes high-resolution 2D panoramic renderings of each scene and their associated point cloud features. Designed to reflect the complexity and variability of construction site data and released to address the shortage of real-world data for geometric processing in construction applications, Rohbau3D facilitates research into scene understanding and computer vision for structural and civil engineering. To our knowledge, it is the first dataset of its kind and scale to be published. Rohbau3D is designed as a foundation for ongoing work, with intention to extend it through additional targeted annotations and as a benchmark to support future research.
我们介绍了Rohbau3D，这是一个新颖的三维点云数据集，逼真地呈现了室内建筑环境。该数据集包含504次高分辨率激光雷达扫描数据，是通过地面激光扫描仪在14个不同的建筑工地采集的，这些工地均来自高层建筑，处于不同的建筑外壳施工阶段或翻新阶段。每个点云都富含激光反射强度标量、RGB颜色值以及重建的表面法向量。除了三维数据，该数据集还包括每个场景的高分辨率二维全景渲染图及其相关的点云特征。Rohbau3D旨在反映建筑工地数据的复杂性和多样性，并发布以解决建筑应用中几何处理的真实世界数据短缺问题，它有助于开展结构和土木工程领域的场景理解和计算机视觉研究。据我们所知，这是同类首个且规模如此大的已发布数据集。Rohbau3D被设计为后续工作的基础，旨在通过额外的定向注释对其进行扩展，并作为支持未来研究的基准。
