Point2Building: Reconstructing Buildings from Airborne LiDAR Point Clouds

Yujia Liu,Anton Obukhov,Jan Dirk Wegner,Konrad Schindler

2024-07-29

Abstract:We present a learning-based approach to reconstruct buildings as 3D polygonal meshes from airborne LiDAR point clouds. What makes 3D building reconstruction from airborne LiDAR hard is the large diversity of building designs and especially roof shapes, the low and varying point density across the scene, and the often incomplete coverage of building facades due to occlusions by vegetation or to the viewing angle of the sensor. To cope with the diversity of shapes and inhomogeneous and incomplete object coverage, we introduce a generative model that directly predicts 3D polygonal meshes from input point clouds. Our autoregressive model, called Point2Building, iteratively builds up the mesh by generating sequences of vertices and faces. This approach enables our model to adapt flexibly to diverse geometries and building structures. Unlike many existing methods that rely heavily on pre-processing steps like exhaustive plane detection, our model learns directly from the point cloud data, thereby reducing error propagation and increasing the fidelity of the reconstruction. We experimentally validate our method on a collection of airborne LiDAR data of Zurich, Berlin and Tallinn. Our method shows good generalization to diverse urban styles.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The paper aims to address the problem of reconstructing 3D buildings from airborne LiDAR point clouds. Specifically, the diversity in building designs (especially roof shapes), low and uneven point density in the scene, and incomplete coverage of building facades due to vegetation occlusion or sensor perspective make it challenging to reconstruct 3D buildings from airborne LiDAR data. To tackle these issues, the authors introduce a generative model called Point2Building, which can directly predict 3D polygonal meshes from the input point cloud.

Unlike many methods that rely on preprocessing steps such as exhaustive plane detection, the Point2Building model learns directly from the point cloud data, thereby reducing error propagation and improving reconstruction fidelity. Experimental validation shows that this method has good generalization capabilities on airborne LiDAR datasets from Zurich, Berlin, and Tallinn, making it suitable for diverse urban styles. In summary, the goal of this study is to efficiently reconstruct high-quality 3D building models from complex point cloud data.