City3D: Large-Scale Building Reconstruction from Airborne LiDAR Point Clouds

We present a fully automatic approach for reconstructing compact 3D building models from large-scale airborne point clouds. A major challenge of urban reconstruction LiDAR clouds lies in that the vertical walls are typically missing. Based on observation buildings consist planar roofs connected with to ground, we propose an infer directly data. With segments both and walls, hypothesize faces surface, final model is obtained by using extended hypothesis-and-selection-based polygonal surface framework. Specifically, introduce new energy term encourage roof preferences two additional hard constraints into optimization step ensure correct topology enhance detail recovery. Experiments various have demonstrated method superior state-of-the-art methods terms accuracy robustness. In addition, generated dataset our consisting 20k real-world buildings. believe this can stimulate research use city applications.