Data Fusion towards building roof reconstruction based on CSR approach

Abstract

3D building modeling has emerged as a focus of research as well as development for many urban planning and geo-information related applications. To its most abstract form on geometric level, 3D building roofs are regarded as the fundamental elements to present outlines of 3D building models. The emphasis of this study is to demonstrate a novel algorithm of building roof reconstruction, termed CSR (Construct-Shape-Refine). The proposed algorithm aims at reconstructing building roof models purely by employing 3D line features or integrating 3D and 2D line features from existing data. In particular, the latter integration involving fusion from multiple data sources strengthens the reliability as well as the precision of 3D building roofs. The CSR approach is characterized through three stages. Geometric inferences are imposed at the stage of Construct where topological relationship of the 3D line features is established while 3D coordinates of roof corners are adjusted and estimated through Shape processes, apart from compensating the hidden boundaries, if any. The efficiencies of two fusion streams, including fusion of 3D line features that result from different sources or fusing 3D line features with photogrammetric line features, are gained as promoting the building roof quality via Refine workflow. Experiments of the Construct-Shape stages show that the proposed method is independent of building roof types, however constrained in polygons under current consideration, as well as whether 3D line features are complete for both successful and efficient performance of building roof reconstruction on a 3D line feature basis. Moreover, the Refine stage offers more up-to-date and satisfactory building roof reconstruction results as compared to the situation when only one single data set is considered.