PATH-TRACING SEMANTIC NETWORKS TO INTERPRET CHANGES IN SEMANTIC 3D CITY MODELS

Abstract

Recent years have seen a significant and rapid rise in the development and deployment of many urban digital twins worldwide. Urban digital twins provide a central platform for incorporating data and knowledge from different sources and fields, and can thus be used for many urban-related processes such as urban planning, analysis, monitoring and visualization. One of the key requirements of digital twins in general and urban digital twins in particular is the continuous, bidirectional data flow between the physical entity and its digital counterpart. Consequently, this requires the ability to both detect and understand changes between different temporal versions of the physical and digital entities. In the context of smart cities and semantic 3D city models however, only a few studies have addressed this so far. This is due to the facts that: (1) Semantic 3D city models (mostly encoded in CityGML) contain multifaceted information modelled in a complex inheritance hierarchy, which complicates their change detection process; (2) As cities constantly evolve over time, matching their datasets often results in a large number of changes, which must be further analysed and processed to produce meaningful information; (3) Individual changes in the datasets are multi-layered as well as often correlated and cannot be fully understood without considering their context in the city model; and (4) Different types of changes are perceived and interpreted differently by different stakeholders, meaning that the outcome of the change detection and interpretation process must ultimately serve the human factor. Therefore, to address these challenges, this research proposes a Path-tracing Semantic Network (PSN) to interpret detected changes in semantic 3D city models. The framework represents the multidimensional nature of changes together with stakeholders in a semantic network, where their interrelations can be analysed explicitly using graph-based path-tracing methods.