Optimization of automatied retrieval of semantic 3D city data

Abstract

A 3D city model is considered as the digital representation of a city/ urban area that may decompose into its objects/ elements such as buildings, roads, railways, terrain, water, vegetation etc., with clearly defined semantics, spatial and thematic properties. Depending on the level of detail (LoD), these objects may further decompose into more detailed features. The OGC standard CityGML, optimally allows integration of the diversified geoinformation of the aforementioned elements and provides multiple resolution at different LoDs. Since 2008, it has been an international OGC standard for representing and exchanging a 3D city model while in 2012, version 2.0 of this standard was published. CityGML represents the geometrical, semantic, and visual aspects of 3D city models and, for this reason, it is considered as an optimal standard for the representation of 3D city models. However, the structure of the CityGML standard is rather complex in order to support all these capabilities. Initially, CityGML was designed for the representation of 3D city models and not for presenting or visualizing 3D city models directly on the web. Therefore, the retrieval of the available semantic features from this standard, by implementing interoperable approaches without the need for specific knowledge, is a challenge, thus constituting the main research question of this work. Achieving this CityGML data retrieval is structured on the basis of interoperability, easy-to-use, semantics and non-expert user.

The current dissertation is structured in six chapters in order to address the research question raised above and the resulting sub-questions. First, the available research works and studies focusing on retrieving CityGML data are examined. Then, the solution of the REST approach is presented and compared with other state-of-the-art technologies, and finally, the CityGML RESTful Web service is conceptually designed and presented as a new approach for retrieving CityGML data based on their semantic characteristics.

Chapter 2 presents the relevant research work that focuses on the CityGML data retrieving utilizing tile or hierarchical-based or Web service-based approaches. Initially, the file-based formats such as X3D, JSON, KML and glTF have been further studied. Next, the OGC I3S and OGC 3D tiles are further examined as they provide a good solution in relation to the literature research. Next, taking into account the complex structure of the CityGML standard and the need to retrieve data from distributed sources, the adoption of the available OGC Geospatial Web services are examined, such as OGC 3DPS and OGC WFS. Also, the extension of the OGC WFS, as well as the integration of the RESTful service architecture on top of OGC WFS are further examined.

The third Chapter of this dissertation studies the interoperable and easy-to-use information retrieval of CityGML based on its semantic characteristics using non-OGC Web services, such as SOAP and REST. Additionally, the REST is further compared with new state-of-the-art technologies that can be adopted as CityGML data retrieval mechanism, such as GraphQL and Falcor. Next, the solution of REST approach is presented and several principles and constraints in respect to the RESTful implementation are described. Thereafter, several principles and guidelines are provided with regard to the CityGML RESTful Web service and finally, the conceptual design of its core resources is presented such as “citymodels” and “gmlid”.

Chapters 4 and 5 focus on the presentation and description of the conceptual design of CityGML RESTful Web service, which is a new approach and proposal of the current dissertation. So, taking into account the CityGML architecture, the CityGML structure is more semantic than geometric, and therefore the retrieval of the data has to be achieved mainly in compliance with the CityGML’s semantic information. From the five components of the CityGML’s architecture, only the component of the thematic modules defines the semantic features of CityGML. Therefore, these thematic modules are defined as the main resources of the CityGML RESTful Web service. However, apart from the above-mentioned resources, some extra main resources are also defined to make easier accessing their available semantic features. Since CityGML adopts the multi-scale modelling in five different LoDs, the same object may be simultaneously represented in different LoDs, enabling the analysis and visualization of the same object with regard to different resolution. However, LoD is considered vital not only in the geometric determination of the level of detail, but also in the semantics. By increasing the LoD, the semantic richness of CityGML increases respectively. Therefore, this semantic enrichment of each of the thematic modules is retrieved by implementing a variety of sub-resources. Thus, some of the main resources have LoD-based sub-resources and hence, their semantic retrieval is available based on the LoD, while, some resources are LoD-independent with no differentiation regarding their semantic sub-resources from one LoD to another.

More specifically, the fourth Chapter deals with the conceptual design of the LoD-based thematic resources of the CityGML RESTful Web service. In this direction, the “bldg”, “tun” and “brid” main resources and their respective child resources are presented. These resources refer to the respective building, bridge and tunnel modules of the CityGML 2.0. Additionally, for each of these resources, various case studies using semantic requests are exploited and presented.

The conceptual design of the rest of the main resources of the CityGML RESTful Web service are presented in Chapter 5. These resources are mainly LoD-independent thematic resources and therefore, they are enriched with semantic characteristics either independently of LoD or from LoD2 and above without any different from one level to another.

Finally, Chapter 6 concludes this research work by discussing the findings of the previous chapters and responding to the sub-research questions formulated to address the aim of this dissertation. Suggestions for future research works are discussed, aiming at making this approach an OGC standard, and on upgrading it so that the upcoming version 3 of CityGML can be fully supported.