Structured Urban Reconstruction
1University College London 2Miami University 3KAUST
Conditionally Accepted: SIGGRAPH-Asia 2017
A central problem in urban modeling is the creation of high quality semantically parsed 3D models for densely built areas. Although recent advances in acquisition techniques and processing algorithms have resulted in large scale imagery or 3D polygonal reconstructions, such data are typically noisy, incomplete, and lack any semantic structure. In this paper, we present an automatic data fusion approach that produces high quality semantic models of city blocks. Starting from polygonal meshes, street-level imagery, and GIS ownership footprints, we formulate an integer program that globally balances the different error sources to produce semantically parsed mass models with associated facade elements. We demonstrate our system on data from four different cities of varying complexity with the most complex examples involving densely built urban blocks spanning hundreds of buildings. In our largest example, we produce a semantic model covering nearly 37 blocks with a total of 1000+ buildings at a scale and quality previously almost impossible to achieve automatically.