Structural approach for building reconstruction from a single DSM

Area of Science:

• Geoinformatics
• Computer Vision
• 3D Reconstruction

Background:

• Accurate 3D building reconstruction is crucial for urban planning and analysis.
• Existing methods often struggle with complex structures and dense urban environments.

Purpose of the Study:

• To develop a new approach for reconstructing buildings from a single Digital Surface Model (DSM).
• To enable detailed and accurate 3D representations of complex buildings and urban areas.

Main Methods:

• Treating buildings as assemblages of 3D parametric blocks from a library.
• Extracting 2D-supports for urban structures (interactively or automatically).
• Utilizing a Gibbs model for 3D block placement and fitting, with Bayesian decision and Markov Chain Monte Carlo sampling for optimization.

Main Results:

• Successful validation on diverse datasets, including satellite (0.7m) and aerial (0.1m) DSMs.
• Generation of detailed 3D representations for complex buildings and dense urban areas.
• Demonstrated effectiveness across a wide range of data resolutions.

Conclusions:

• The proposed method offers a robust and versatile approach to 3D building reconstruction from single DSMs.
• The LEGO-like block assemblage strategy effectively handles complex urban structures.
• The technique provides high-quality 3D city models adaptable to various levels of detail.