Towards mainstream 3D geographical data

Jantien Stoter, Jacob Beetz, Hugo Ledoux, Marcel Reuver, Rick Klooster, Paul Janssen and Friso Penninga
[email protected]

Introduction
Over the past ten years technologies for generating, maintaining and using 3D geo-information have matured. In addition many local governments have 3D models of the city, a large number of companies are providing services for constructing 3D models, and universities and research organisations are investigating 3D technologies (3D re-construction, data management, validation and visualisation). Yet many (governmental) organisations face numerous challenges in introducing 3D applications and technologies in their day-to-day processes. Despite the practical difficulties, it is clear that 3D information is becoming increasingly important in many applications. These developments motivated a pilot in the Netherlands to advance the use of 3D in this country. The pilot was initiated by the Dutch Kadaster, Geonovum (the National Spatial Data Infrastructure executive committee in the Netherlands which develops and manages the geo-standards), the Netherlands Geodetic Commission (NCG) and the Dutch Ministry of Infrastructure and Environment.

From January 2010 until June 2011 a uniform approach for acquiring, maintaining and disseminating 3D geo-information has been explored in a collaboration between over 65 stakeholders in The Netherlands (Stoter et al. 2011). A major result of the pilot was the proof of concept for a 3D Spatial Data Infrastructure (SDI), covering issues on the acquisition, standardisation, storage and use of 3D data. The findings of the pilot were formally established in a national 3D standard realised as a CityGML Application Domain Extension. The ADE completely integrates the OGC CityGML Encoding Standard (OGC, 2008) with a new version of the existing national Information Model for Geo-information (called IMGeo; described in IMGeo 2007). IMGeo contains object definitions for large scale representations of roads, water, land use/land cover, bridges, tunnels etc. and prescribes 2D point, curve or surface geometry for all objects. As the new version of IMGeo is completely integrated with CityGML, (see Figure 1), IMGeo version 2.0 also facilitates extensions to 2.5D representations (i.e. as height surfaces; equivalent to CityGML LOD0) and 3D (i.e. volumetric; i.e. CityGML LOD1, LOD2 and LOD3) representations of the objects according to geometric and semantic principles of CityGML.

Figure 1: TunnelPart AD Element with 2D geometry

The close integration between an existing information model for 2D geo-information and CityGML is an important step towards the practical use and re-use of 2D and 3D information. Further technical details about the ADE are reported in Van den Brink (2011).

Although the 3D standard is an important prerequisite for further 3D developments, wide use of 3D is still not common practice in the Netherlands. Further advances are required to assure that 3D Pilot results are implemented in actual applications. Therefore a follow-up pilot has been started focusing on further research within a similar collaborative and experimental environment as the first phase of the 3D Pilot. Section 2 explains the set up of this follow-up project and details the six activities. Initial conclusions and work in progress are finally described in Section 3.

The use of 3D data in applications was studied during the first phase of the 3D Pilot and it is therefore not further explained in this paper. Demonstrations of the use cases can be found at Geonovum (2012c).

3D Pilot NL phase II
In the development process of CityGML ADE IMGeo 2.0 a number of topics were identified that requires further attention before the standard can be widely implemented, e.g. how does the national 3D standard works in practice including the consequences of this new modelling method for IMGeo when used for both 2D and 3D datasets, e.g. how to generate compliant data; how to preserve the links between the different Levels of Detail (LODs); and how to upgrade 2D LOD to higher LODs. These open issues are currently being studied in a follow-up project of the 3D Pilot.

The goal of the follow-up pilot is more focused than the first pilot and aims at writing best practice documents by joint effort of the 3D Pilot community. The best practice documents are based on tools and techniques that are being developed for supporting the implementation of the 3D standard. Specific attention is being paid how to align CityGML to the standard in the BIM (Building information Model) domain (IFC).

In summer 2011 a new call was launched responded by over 100 organisations. These organisations, listed at Geonovum (2012b), are currently executing the six activities of the second 3D Pilot NL. The activities, including background, motivations and work in progress, will be further explained in the remainder of this section and are:

• Generating example 3D IMGeo data for several levels of detail and classes
• Writing example tendering documents for creating 3D information
• Designing and implementing a 3D validator
• Describing a generic approach for maintenance, update and dissemination of 3D IMGeo data
• Collecting examples of 3D killer applications
• Align CityGML and IFC/BIM

Generating example 3D IMGeo data
To understand how IMGeo works for the integrated 2D and 3D approach, example 3D IMGeo data is being built and made available to wider audiences. The example data is also used to check whether CityGML compliant software is capable to understand the 3D IMGeo data. The example data will be specifically useful to:

• Obtain insights into the 3D aspect of our approach including different LOD’s, i.e. for buildings the LOD concept is well
• defined, but how does the LOD concept applies to other objects such as trees?
• Provide the possibility for third (new) parties to experiment with 3D IMGeo data;
• Provide test data for 3D validation tests (see section 2.3);
• Show how the standard is interpreted when applied to real data (helpful for future data providers).

The source data that has been made available for the test area on the 3D Pilot data server (hosted by the Delft University of Technology) are:

• IMGeo compliant 2D data (see Figure 2a), provided by the municipality of Den Bosch;
• Stereo photos (30 march, 2011), 10 cm resolution, provided by the municipality of Den Bosch;
• Point cloud (3 points per m2), DTM&DHM, date: April 2009, provided by the municipality of Den Bosch;
• High resolution laser data (selected from a data set available for the whole country: Actueel Hoogtebestand Nederland, AHN), provided by Het Waterschapshuis;
• Ortho photos, provided by Cyclomedia;
• High resolution point cloud obtained from terrestrial laser scanning, provided by Cobra, see Figure 2b;
• Point clouds generated from aerial photographs, provided by Imagem;
• Oblique photos, provided by Slagboom en Peeters.

Figure 2: example source data

To get thorough insight into the key aspects of 3D IMGeo data including how the LOD concept applies to several themes, the 3D Pilot participants currently extend 2D IMGeo data into 2.5D and 3D according to geometric and semantic principles of CityGML (see Figure 3).

Figure 3: Objects in test area modelled in CityGML LOD3

Writing tendering documents for creating 3D information
Usually a municipality will outsource the 3D data acquisition for 3D IMGeo data. For most municipalities 3D is a new domain and example tendering documents may help them to precisely specify what to ask the market. In a next step, when the data is delivered, the specifications can be used as acceptance criteria, i.e. to check the quality of the data. For private companies these documents are helpful since they both clarify and unify the demand for their services.

Apart from the experiences gained from building example data (activity 1), the tendering documents will be based on experiences of cities that have already invested in 3D city models, i.e. The Hague and Rotterdam. Both cities faced difficulties in comparing offers from different companies because the specifications appeared to be interpretable in several ways and this also caused problems in setting up acceptance criteria for the delivered product. The result is that the CityGML datasets differ between the two cities but it is not always clear whether this was intended.

Since the example tendering documents will be a joint effort of the 3D pilot community, they will be based on knowledge, interests and experiences of research institutes, private and governmental organisations and not only based on the information available at the bidder as currently practiced.

Several variants of the tendering documents are possible based on the available source data – e.g. point clouds (either obtained by airborne or terrestrial laser scanning), high resolution aerial photographs, oblique photos – and the ambition level (i.e. which information at which LOD).

Design and implementation of a 3D validator
A validator is necessary as an independent tool to verify whether a dataset is compliant with IMGeo 2.0, or not. This also applies for the 3D extensions. When validating objects, it is necessary to validate both the semantics and the geometry. The former is according to the classes of CityGML and/or of the IMGeo extensions, and the latter is according to the international specifications (e.g. ISO19107 and GML). Geonovum has already built a validator for IMGeo datasets (the software is available as open source software, see Geonovum (2012)), but it is only for two-dimensional primitives.

This activity primarily studies which functionalities are necessary to validate the geometry of 3D solids. During the first 3D Pilot we have noticed that several real-world datasets have objects that appear to be valid when looking quickly at them, but in reality they are not. Figure 4 shows two examples. These (often small) problems prevent users from, for instance, convert their objects to other formats (including BIM and CAD formats, see Section 2.6) and also to analyse them (the volume of an invalid solid could be impossible to calculate).

Figure 4: Two real-world invalid buildings.

While different definitions of a valid 3D object are used in different disciplines, we focus on the definition given in the ISO standards (ISO 2003) and implemented with GML (OGC 2007). A GML Solid: “The extent of a solid is defined by the boundary surfaces as specified in ISO 19107:2003. gml:exterior specifies the outer boundary, gml:interior the inner boundary of the solid” (OGC, 2007). Without going into all the details, we can state that a solid is represented by its boundaries (surfaces), and that like its counterpart in 2D (the polygon), a solid can have ‘holes’ (inner shells, or cavities) that are allowed to touch each others, or the outer boundary, under certain circumstances. To be considered a valid solid, a solid must fulfil several properties. The most important are: (i) it must be simple (no self-intersection of its boundary); (ii) it must be closed, or ‘watertight’; (iii) its interior must be connected; (iv) its boundary surfaces must be properly oriented; (v) its surfaces are not allowed to overlap each other.

We are currently building an open-source 3D validator. This is because none of the surveyed GIS packages that provide functionalities for validating 3D objects was fully compliant with the definition of the ISO. Our validator is based on the work of Ledoux et al. (2009) and is ISO-compliant. It uses advanced data structures and operations to analyse the topological relationships between 3D objects. Furthermore, it will be built as an extension to the current validator for 2D (developed by Geonovum) so that the geometry of 3D objects can be taken into account while using the same website with the same workflow.

Finally, other validation issues will be investigated. We plan to investigate the validation of not only solids, but also 3D MultiSurfaces as these are often used (buildings are often modelled without the ground floor for instance).

Maintenance, update and dissemination of 3D IMGeo
After having invested in a good 3D model, the next question is how to maintain and update the model. Can mainstream DBMSs be used? How to update: integrated with the existing data processes, renewed creation when the 2D data changes or a mix of both? For the maintenance of the data it is a relevant question how to guarantee that 3D IMGeo data remains synchronized with the 2D data. The challenge differs if the 3D data is managed separately from 2D (how to maintain 3D data? In a 3D spatial DBMS?) or generated on the fly.

An important first step is to obtain more insight into how CityGML data encoded in CityGML files can be maintained and updated.

Therefore a challenge was organized in order to study the state-of-the-art of 3D editing in commercial software. Four neighboring CityGML data sets (courtesy of the Municipality of The Hague) were provided and the following challenges were defined:

In addition to these challenges it was mentioned that it was up to the vendors to decide in which environment or format the actual edits were made, as long as both input and output were in CityGML format without any loss of data. Up till now the following companies demonstrated their capabilities: StrateGis (see Figure 5), Toposcopie, CPA Systems, Safe and Bentley.

Figure 5: Screenshots of the StrateGIS solutions

The preliminary conclusion of the challenge to maintain CityGML data is first that the five vendors showed solutions that (partially) rely on either Google SketchUp (or the Google SketchUp API) or FME. In addition database solutions for 3D data are rare. Therefore the availability of good import and export functionalities for CityGML (and the ADEs) is essential, which gave motivation to plan a “CityGML relay” as follow up step of these challenge-outcomes (work in progress).

Collecting examples of 3D killer applications
Although 3D applications is common practice for many professionals, 3D is new and considered as “complex” and “expensive” to others. To show the need for 3D to policy makers and new comers in the field, this activity is collecting examples of 3D applications that are already practised by the 3D pilot participants and make them available in an easy accessible format (flyer, PPT, Website). Specific attention is paid to the integration-role of 3D information, i.e. as base information for many applications, see Figure 6.

Figure 6: Screenshot of the 3D Pilot Website that collects the killer applications (the circles represent the different uses cases)

Align CityGML and IFC/BIM
In both GIS and BIM domains it is acknowledged that the integration of both types of data is beneficial. BIM data is commonly modelled in the IFC standard and 3D GIS data can be encoded in CityGML. The two standards model similar object types. Therefore it is relevant to see how these two standards map, integrate and interact with each other.

BIM (i.e. design) data can feed GIS data and GIS can serve as reference for BIM data. However, integration should acknowledge the differences between both types of data. To start with, the object description of BIM and GIS (e.g. CityGML LOD4) differs significantly. In addition GIS is characterised by coverage of large areas (e.g. a complete city) and lower precision, while BIM is characterised by its local and very detailed approach, the limited number of construction models usually available in a city and high precision necessary for reliable construction calculations. Also the modeling approaches of CityGML and IFC differ significantly, i.e. IFC models much more classes and allow also non-hierarchal relationships, where CityGML contains a limited number of classes structured via hierarchical relationships. Another core issue for bidirectional transformations are additional geometry types that are handled in the building industry and can be captured in IFC instances (Nagel, 2007). Among them are Boundary Edge Representations (BRep) and Constructive Solid Geometry (CSG), which are frequently used as implicit capturing formats while CityGML is limited to explicit polygonal representations. While polygonal representations can be derived from these geometry types in a straightforward manner (thus transforming IFC to CityGML), it is impossible to generate e.g. efficient CSGs from triangulated surface representations.

Several studies have shown that a conversion between IFC and CityGML is possible, see (Isikdag and Zlatanova 2009, Berlo and De Laat 2010; Bormann 2010; El-Mekawy 2010). However, because of the different modelling approach of both information models, there is not one optimal nor uniform conversion.

Therefore, based on experiments and a study on best practices, this activity is working on making agreements how to best realise the alignment between the two standards.

For example, agreements on a unique mapping between IFC and CityGML to make sure that a conversion always happens in the same meaningful way. This will also avoid the currently common situation that the rich semantics of IFC is lost because all objects are converted in the GenericObjectClass. Also it may help to model according to specific rules in IFC to make sure that specific CityGML concepts can be derived (e.g. LOD2 Buildings) from the IFC data. Those agreements will be formulated as recommendations to the relevant standardisation organisations, i.e. as change requests to BuildingSmart (2012) and OGC for generic issues and to national standardisation organisations for the national specific issues.

Because IFC and CityGML both serve different applications, it is important that both the original IFC source data and a CityGML representation are available and that CityGML objects explicitly refer to their interrelated IFC objects and vice versa. In this specific activity we study how this can be done by joint effort from IFC and CityGML experts.

Initial conclusions and work in progress
This paper presents the follow-up of the 3D Pilot NL, which is a large collaboration in the Netherlands aiming at pushing 3D developments in the Netherlands. The first phase resulted in a national 3D standard. Some results and insights obtained during the first phase are sufficiently mature to be anchored in practice such as maintaining and further developing the 3D standard by Geonovum and the provision of a countrywide 3D midscale base dataset which is currently under study at the Kadaster (collaboration with University of Twente). Other results of the first 3D Pilot NL phase need further attention, specifically how the new 3D standard works in practice. This is currently being further explored in a second phase of the 3D Pilot in which 100 organizations are participating.

The main conclusion of running the 3D Pilot is the change of vision concerning 3D in the Netherlands. At the start of the 3D Pilot (March, 2010) many saw that 3D had potentials, but did not know how to deal with 3D. In the course of the pilot the ambitions for 3D have become much more focused, also supported by the national 3D standard. These ambitions are further developed now the second phase of the pilot is running. Several aspects appear to be crucial for the adoption of the 3D standard. Firstly, the engagement of many stakeholders is important to gain the necessary support. Secondly, aligning to the ongoing 2D efforts makes that 3D applications become in reach for governmental organizations. In addition, collaborating is important because the issues of 3D are complex and sharing knowledge between different 3D experts is therefore important to realize innovations. Finally, it has been important that some national organizations took the responsibility to facilitate the process. Although the pilot is a joint effort and ‘owned’ by the community, national organizations have to initiate and facilitate such a network organization and they are important for anchoring the results.

Currently the work on the six activities of the 3D Pilot NL II is running in parallel, supported by discussions within the 3D Pilot NL LinkedIn group (over 500 members) and frequent meetings. 3D test data have been prepared for the test area and several participants are currently working on generating different LODs and different themes for 3D IMGeo data. The 3D validator is being developed, a contest for maintaining and updating 3D CityGML data has been launched and killer applications for 3D are being collected. In addition the information models IFC and CityGML are studied for possible integration, and the possible mappings, alignments and conversions are discussed in dedicated working sessions. Also the integration of 3D IMGeo with the subsoil (i.e. geology and cables & pipelines) is being studied (see also the work of Becker et al 2010, Hijazi 2010, Zobl and Marschallinger, 2008).

The 3D Pilot will finish in summer 2012. Among the end results are: examples of 3D IMGeo data, a 3D validator, best practice documents of how to acquire, maintain, update and disseminate 3D IMGeo data, demonstrators that show the potentials of 3D, and recommendations for further developing CityGML compatible with 3D standards in other domains and with the established 2D information models. The results will be presented to the wider (professional) public in a special issue of a Dutch professional magazine on geo-information and a national 3D symposium.

From our national pilot we have observed that 3D is increasingly vital for managing and planning our densely built environment. Therefore 3D information will be more and more important for governmental organisations. To move forward in the highly complex domain of 3D information, we consider agreements and collaborations essential. In addition a national consensus on a generic 3D approach supported by a 3D standard diminishes the risks of investment for individual organisations. This is accomplished in the 3D Pilot NL.

Acknowledgements
We would like to acknowledge the contributions of all 3D Pilot participants which is very important for the achievements in our country. This research is supported by the Dutch Technology Foundation STW, which is part of the Netherlands Organisation for Scientific Research (NWO) and partly funded by the Ministry of Economic Affairs, Agriculture and Innovation (project code: 11300).

References

• Becker, T., C. Nagel and T.H. Kolbe, 2010. Integrated 3D modeling of multi-utility networks and their interdependencies for critical infrastructure analysis, 5th International 3D GeoInfo Conference, November 2010, Berlin
• Bormann, A., 2010. From GIS to BIM and back again – A Spatial Query Language for 3D building models and 3D city models, 5th International 3D GeoInfo Conference, November 2010, Berlin
• BuildingSmart, 2012. [online] available from www.buildingsmart.nl/ Clement, J, 2011. Presented at the second meeting of 3D Pilot Phase II [online} available from
• El-Mekawy, M.S., A. Östman, K. Shahzad, 2010: Towards Interoperating CityGML and IFC Building Models: A Unified Model Based Approach, 5th International 3D GeoInfo Conference, November 2010, Berlin
• Geonovum, 2012, [online]
• Geonovum, 2012b. [online] available from
• Geonovum, 2012c. Demonstrations of the 3D Pilot uses cases [online available] www.geonovum.nl/dossiers/3dpilot/bibliotheek/presentaties#films (YouTube).
• Hijazi,I., M. Ehlers, S. Zlatanova, T. Becker, L. van Berlo, 2010: Initial investigations for modeling interior Utilities within 3D Geo Context: Transforming IFC-interior utility to CityGML/UtilityNetworkADE, 5th International 3D GeoInfo Conference, November 2010, Berlin
• IMGeo, 2007, Informatiemodel Geografie, oktober 2007 [online]. www.geonovum.nl/sites/default/files/IMGeo_rapport_definitief_versie_1.0.pdf
• Isikdag, U. and S. Zlatanova, 2009: Towards defining a framework for automatic generation of buildings in CityGML using Building Information Models In: J. Lee and S. Zlatanova (Eds.); 3D Geo-Information Sciences, Springer, 2009, pp. 79-96
• IS0, 2003, ISO 19107:2003 Geographic information — Spatial schema, [online] Available from: www.iso.org/iso/catalogue_detail.htm?csnumber=26012
• Ledoux,H., E. Verbree and H. Si, Geomatric Validation of GML Solids with the Constrained Delaunay Tetrahedralization, in: De Mayeer, Neutens, De Rijck (Eds.); Proceedings of the 4th International Workshop on 3D-Geo- Information, November 2009, Ghent, pp. 143-148
• Nagel, C. 2007. Ableitung verschiedener Detallierungsstufen von IFC Gebäudemodellen (Derivation of different levels of detail from IFC building models). Master Thesis. Karlsruhe University of Applied Sciences, Germany
• OGC, 2007, OpenGIS® Geography Markup Language (GML) Encoding Standard. Version 3.2.1, doc # OGC 07-036, https://portal.opengeospatial.org/files/?artifact_id=20509
• OGC, 2008, OpenGIS® City Geography Markup Language (CityGML) Encoding Standard, version 1.0.0, document # 08-007r1. https://portal.opengeospatial.org/files/?artifact_id=28802
• Stoter, J., Vosselman, G., Goos, J., Zlatanova, S., Verbree, E., Klooster, R. and Reuvers, M., 2011. Towards a National 3D Spatial Data Infrastructure: Case of The Netherlands. PFG Photogrammetrie, Fernerkundung, Geoinformation, 2011(6): 405-420.
• Stoter, J. , T. Tijssen, E. Verbree and S. Zlatanova, 2011b, Het 3D-testbed van de 3D-Pilot: Ceci n’est pas CityGML (in Dutch), In: Geo-Info, Volume 8, 5, 2011, pp. 19-22
• Van den Brink, L., J.E. Stoter and S. Zlatanova, 2012. Establishing a national standard compliant to CityGML, accepted for International Journal of Geographical Information Science IJGIS
• Van den Brink, L., J.E. Stoter and S. Zlatanova, 2012b. modeling an application domain extension of CityGML in UML, proceedings 3D GeoInfo Symposium, Quebec, Canada, 16-17 May.
• Zobl, F., and R. Marschallinger, 2008. Subsurface GeoBuilding Information Modelling GeoBIM, GEOinformatics, 8, volume 11, December 2008, p40-43

Note: This paper was presented at Geospatial World Forum 2012 in Amsterdam, the Netherlands.