Out-Of-Core Topologically Constrained Simplification for City Modeling from Digital Surface Models
Abstract
We present a framework for rapid reconstruction of building models from very large, high-detail digital surface models (DSM) of urban areas. Our method is based on a geometric mesh simplification approach augmented with shape constraints (Wahl et al., 2008). This approach allows to abstract the full-featured DSM in such a way that important structural elements are maintained irrespective of the approximation accuracy. In this paper we present two major extensions. Firstly, we deal with situations, where the original approach may generate artefacts due to incomplete or inconsistent structural information, mainly caused by vegetation close to the facades. We present refined topological constraints, which handle these problems and a filtering which neglects structural information that contradicts the mesh topology. Secondly, we extend the computational framework to be fully out-of-core capable and present a way to parallelize computations on multiple cores. We demonstrate the efficiency of our method by showing results for downtown Berlin a dataset containing more than 1 billion height samples processed in less than 30 hours.
Keywords: Abstraction, City Model, DSM, Geometry Simplification, Visualization
Images
![]() | ![]() |
Download Paper
Bibtex
@ARTICLE{moeser-2009-topologically, author = {M{\"o}ser, Sebastian and Wahl, Roland and Klein, Reinhard}, title = {Out-Of-Core Topologically Constrained Simplification for City Modeling from Digital Surface Models}, journal = {International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences}, volume = {XXXVIII-5/W1}, year = {2009}, month = feb, keywords = {Abstraction, City Model, DSM, Geometry Simplification, Visualization}, abstract = {We present a framework for rapid reconstruction of building models from very large, high-detail digital surface models (DSM) of urban areas. Our method is based on a geometric mesh simplification approach augmented with shape constraints (Wahl et al., 2008). This approach allows to abstract the full-featured DSM in such a way that important structural elements are maintained irrespective of the approximation accuracy. In this paper we present two major extensions. Firstly, we deal with situations, where the original approach may generate artefacts due to incomplete or inconsistent structural information, mainly caused by vegetation close to the facades. We present refined topological constraints, which handle these problems and a filtering which neglects structural information that contradicts the mesh topology. Secondly, we extend the computational framework to be fully out-of-core capable and present a way to parallelize computations on multiple cores. We demonstrate the efficiency of our method by showing results for downtown Berlin a dataset containing more than 1 billion height samples processed in less than 30 hours.}, issn = {1682-1777} }