Construction method for topological integration of multi-scale datasets supported by a primal-dual data structure

Spatial datasets covering the same area often exist as separate layers, which differ in scale and source. The spatial relations between matching features located at different layers can be retrieved through a computationally demanding process based on geometry and other properties. When these layers are stored without structural links, they may suffer from redundancy and inconsistent updates, making it difficult to identify spatial relationships across representations. The CityGML concept or national cadastral data, when compared with buildings from OpenStreetMap, often show geometrical and semantic inconsistencies. This study introduces a method for the topological integration of spatial datasets, enabling the construction of a full 3D vario-scale representation from 2D individual multi-scale layers. This approach first generates intermediate shared partitions using a geometric overlay. This partition allows us to detect matching feature components and create topological relationships between them across layers. The integrated model is implemented using a primal–dual data structure, where the primal space stores the geometric integrated representation and the dual space maintains the hierarchical-topological relational structure between the corresponding feature instances. Further optimization introduces deep integration and reduction of geometrical redundancy by removing horizontal faces between feature components and merging vertical coplanar faces and colinear edges. The case studies included a mock and also real datasets: administrative boundaries and independent multisource building footprints available in a national database and open repositories. The resulting structure forms a general foundation for future spatiotemporal data modeling and integration of 3D models into a 4D representation.