3D GeoInfo 2012 Conference Summary

Here's parts of my talks notes of the 3D GeoInfo 2012 conference. Hope some elements are useful to some of you.

General notes

  • The whole conference was more of the research / academics type than the commercial providers one
  • "Glocal": global and local at the same time
  • General decline in conference attendance: different ways of networking and training, economic context
  • Next 3D GeoInfo will be in Istanbul in November 2013

Talk notes


Opening talk

Opening talk, Jacynthe Pouliot

  • 7th 3D GeoInfo conference
  • First time in North America
  • The 7th 3D GeoInfo was part of the Global Geospatial Conference 2012 (13th GSDI), which had about 700 participants, vs about 60 specifically for the 3D GeoInfo track

    • Thanks to the global conference, the exhibition hall was pretty big had lots of booths
  • The opening talk concluded with a speech by University Laval faculty dean

Modeling 3D topographic space against indoor navigation requirements

Brown, Germany

  • Requirements for indoor navigation for users
  • Use cases

    • Use cases such as routing to single or multiple destinations
    • Scenarios, considering access controls to locations (rooms)
    • Considering obstacles
  • Existing models: CityGML, IFC, 3d graphics formats... All have severe limitations for indoor navigation (none support navigation constraints)
  • Prohibited maneuver: physically possible, but prohibited, such as a turn on a one way street
  • Semantic and navigation constraints models for indoor navigation
  • Future work to be on extending existing model to remove current limitations for indoor navigation

Indoor localization using wifi fingerprinting for LBS applications

Sohn, York University

  • Augmented city, virtual and physical spaces become part of the same continuum, not two separate dimensions anymore
  • Using the Google Earth plugin for visualization
  • Linking real life webcams with live virtual world models, pretty interesting
  • Wifi fingerprinting requires multiple access points to enhance location accuracy
  • Using both wifi fingerprinting and wifi trilateration

    • Using Received Signal Strength (RSS)
  • Went Technical talk on location accuracies

    • Correctness rates of 80-90%

Enhancing visibility of labels in 3D navigation maps

Vaaraniemi, BMW

  • Visual presentation
  • Labels in 3D are often a problem when they're behind objects
  • Solutions includes transparency, wire framing
  • Goals: readability, mapping to element, spatial orientation
  • Draw behind and draw before doesn't work
  • Semi-transparent approaches
  • Five concepts: baseline, cut-aways (bad!), transparency aura (confusing too), glowing labels, glowing roads (best in my opinion)
  • They implemented prototypes and benchmarking, in OpenGL 3
  • During questions: how does it scale in mobile devices (instead of just cars)
  • During questions: labels at different scales, how does it deal with generalization
  • During questions: research regarding bending the faraway part of the map to better show labels

Semantic 3D modeling of multi-utility networks

Becker, Germany

Missed most of the talk.

  • Energy demands
  • Need: homogeneous representation, 3D topological and functional model, support for hierarchies and interdependencies

Integrating scale and space in 3d city models

Ledoux, Netherlands

  • 5D: space plus time plus scale

    • Integrated into a 5D hypercube
  • Hypercube advantages: consistency, queries, data management
  • Generalizations processes
  • With the cube, there are no precomputed scales (18 for Google Maps and others), the cube is sliced at the requested scale each time
  • Slicing the space-scale cube non-horizontally in the scale

    • Interesting, automatically provides details near and low details away
  • Multiple dimensions slicing
  • Problem: gigantic storage capacity required

    • They'll try putting the cube in database

Generalization and viz of 3D building models in CityGML

Ullah Baig, Malaysia

  • Generalization: reduces data volume, complexity and unnecessary details
  • We're at semantic-based generalization for 3D building models... It evolved quite a lot in the past decade
  • Presented generalization techniques and algorithms

Modeling an application domain extension of CityGML in UML

Van Den Brink , Netherlands

  • Extended IMGeo to 3D

    • CityGML was too generic and IMxxx is rich on semantics, but 2D only
  • Data providers in Netherlands will have, by law, to use the IMGeo model

    • Mandatory, with optional elements
  • Application domain extension = ADE
  • They are at the conceptual stage

    • Next step is test the model

Keynote: Registration, detection and classification in point clouds in urban scenes

Ioannis Stavos, city university of new york

  • Lidar can use the differences in intensity of the return of the signal
  • Accuracy of 5mm
  • Point clustering with planar segmentation
  • Inherent problems at borders
  • Can infer lines and polygons from the point clusters
  • Presentation went technical
  • Range-intensity registration
  • Overlaying textures from 2D pictures on the 3D model
  • Doing Fourier transform to distinguish cars from trees within the lidar point cloud
  • Filling hidden parts (from object in front) using repeating patterns
  • Members of his team went to work for Google
  • Accuracy is about 85-90%
  • This was not really a keynote, the topic could have been in a common track... Nothing high level

Volumetric algorithms for single tree delineation towards A fully automated process for the generation of forests

Buecken, Germany

  • Visually impressive
  • Accuracy of over 90%
  • Excellent presentation despite the lack of notes

A service-based concept for camera control in 3D geovirtual environments

Klimke, Germany

  • So many challenges for 3D environments

    • Including heterogeneity, massive storage, updates, accuracy, etc...
  • Thick clients vs thin clients... Thin are required for mobile devices
  • Proposing a camera control service
  • One of the Goals: transfers part of the complexity and computation power required to the thick client instead of the thin client on mobile devices

Representing 3D topography in a DBMS with a star-based data structure


  • Constrained Delaunay tetrahedralisation (TEN)

    • Ten is an analogy to TIN
  • Why TEN: Storage is is simplified, speed up spatial analysis, features are represented, robust implementation, spatial relations between unconnected features explicitly stored
  • Disadvantage of TEN: large database size

    • this talk is about compression of TEN data
  • Attempts were made in the past, but had serious limitations (not topological)
  • They're proposing a powerful 3D data structure
  • Easy to implement in a RDBMS

    • They did it in PostGIS
  • Pretty interesting


Can topological per-culling of faces improve rendering performance of city models in Google Earth?

Ellul, U.K.

  • Removing 3D polygon faces that are hidden from external view and merging those that overlap

    • Saved about 40% of space
    • Display time was about 2x faster
  • Might be significant for mobile devices and for reduced bandwidth situations
  • They're going to use WebGL instead of the Google Earth plugin
  • Google Earth is limited in terms of ingested file size

    • Solution is a tiled-approach... or switch to WebGL!

Revealing the benefits of 3D topology on under-specified geometries in geomorphology

Lowner, Germany

  • Went technical presentation
  • Lack of standard that fits the needs of geomorphology
  • Handling the fuzziness of geomorphological representations

Geometric-semantical consistency validation of CityGML models

Wagner, Germany

  • Pretty technical talk on validating CityGML models
  • Java application with GUI for the CityGML validation and create reports, called CityDoctor

Keynote - Philippe Cantin: building large cities for video games

  • Formerly from Ubisoft
  • CAD, GIS and video games worlds still have a lot to learn from each other
  • The practical components of moving in 3D video games
  • 3D sound spheres and light sources have to be taken into account
  • The main constraint: memory required to load 3D elements... Constantly freeing and loading 3D objects
  • Required collaboration between stakeholders: developer, marketer,, designer, producer, etc
  • Concurrent edition of 3D environments... Like the Hero Engine (game industry)... Boeing did something similar instead of turn-based edition
  • City engine from the University of Zurich
  • Procedural building: instead of creating objects one by one, algorithms build objects such as building based on styles
  • Gaming industry also going in the cloud

Advancing DB4GeO

Kit.edu Breuning, Germany

  • Open source, at least the core
  • Object-oriented DB written in java
  • Seems to have a lot of advanced features
  • Imports exports CityGML data
  • Includes a web-based access using a RESTful architecture
  • Visualization with WebGL
  • Pretty impressive

Glob3 Mobile: an open source framework for designing virtual globes on iOS and Android mobile devices

Santana, Spain

  • It's a virtual globes framework for mobile devices
  • Oriented for iOS and Android, with a common API

    • iOS constraints, Android fragmentation... So nothing is perfect
  • They use a C++ code converter to create java code for their Glob3 engine
  • WebGL scheme
  • Embeddable 3D globe
  • Tiled layer and WMS compliant
  • Showed a demo and seems to work great!
  • Next steps...

    • Html5 and WebGL... Currently runs very well
    • Add 3D models and geometries to the globe
  • Glob3m on glob3.org & Ami.dis.ulpgc.es/glob3m

Reconstruction of rectilinear building facets

Utrecht university, Lankveld

  • Point-based urban reconstruction
  • Technical presentation
  • Making polygons out of a cloud of points in 2D
  • Seems pretty effective

Reconstruction, storage and application of 2.7D models

Gorte, delft university

  • TIN with vertical walls... 2.7D is between 2.5D and 3D
  • Context is urban hydrology
  • 2.5D DEM with 3D buildings
  • Using the star model
  • Storing TINs in a single table
  • Removing points that are coplanar, then removing colinear middle points

A 3D GIS implementation for realizing 3D network analysis and routing simulation for evacuation purposes

Karas, Turkey

  • Evacuation scenarios
  • Using Oracle Spatial and a java-based 3D GIS, JOGL bindings for OpenGL
  • Supports CityGML, they used for visualization
  • Network analysis all done within Oracle Spatial
  • They built a routing instructions engine, instructions for people getting evacuated

3D Geospatial modeling and visualization for marine environment: the Canadian arctic

Sahlin, Canada

  • The biosphere is essential composed of the marine ecosystem, the land-based is of a small volume in comparison
  • Study in the Beaufort Sea area, near the Mackenzie river delta

    • 560km wide delta, quickly goes to 1km depth
  • Looking at snapshots in time, the temporal component is usually essential to consider in oceanography
  • He found that the best tool for him was GoCAD

    • Table comparing the various options
  • Continuous 3D space not well integrated into 3D GIS modeling tools
  • 3D GIS still have a lot of lacks in terms of 3D spatial analysis comparing to the 2D analysis features
  • Kringing interpolation worked best for them
  • They tried EnterVol for ArcGIS

Keynote: Richard Mongeau - measurements and 3D representations of the territory by land surveyors of the City of Montreal

City of Montreal, Geomatics Division

  • Talk in French but with slides in English... pretty strange since most of the audience can't understand French
  • Land surveyors act... Law that specifies reserved acts for land surveyors

    • Recognized by OIQ
  • 90 colleagues at their Geomatics Division
  • 4120 geodetic points on the Montreal island, passive network
  • Active network of about 12 geodetic locations
  • 3D city model built for the 'Montreal 2025' project done almost a decade ago

    • They significantly update it since then
  • With Bentley, LOD-1 and LOD-2, data is piped directly into CityGML standard
  • Data sometimes comes from photogrammetry
  • They use RhinoTerrain software
  • They also use terrestrial lidar
  • Conveying positional accuracy metadata is crucial, users must know what they have in their hands
  • Finding the sources of errors and data integration issues
  • Lasergrammetry with target for calibration and validation
  • Data customers sometimes ask for centimeter accuracy while they don't really need such accuracy
  • Legal cadastre in 3D, underground, such as the new planetarium project
  • They use Bentley software for sewer networks and water distribution networks
  • They use color coding for representing the various categories of accuracy
  • The level of detail they have for the whole city is impressive
  • They develop a lot of tools with Bentley, which certainly is happy to work with such a customer
  • They also do some dynamic data acquisition, such as mobile lidar
  • They have an automated tree inventory
  • They use a lot of various software, Bentley, Oracle Spatial, FME, ArcGIS, 3D PDF, and even Google Earth
  • They have technical standards and even a 3D ethics charter!
  • Apparently, not that many municipalities have so integrated 3D systems
  • Excellent and funny talk, but since it was in French, he lost a lot of attendees

A 3 steps Procedure for enriching augmented reality games with CityGML 3D semantic modeling

Zamyadi, Canada

  • Energy Wars mobile game

    • Location-based scenario and uses augmented reality
    • This was a Geiode-funded project
  • Absence of a methodology for AR-oriented 3D data modeling
  • Their 3 steps: reverse engineering, enriching with 3D features, mapping to CityGML
  • One of the main reason for choosing CityGML, is for interoperability
  • Using a CityGML ADE (application domain extension)
  • This is mostly at the UML stage, the impact on the modeling process has not been tested yet

Implementation of a National 3D standard: Netherlands

Zlatanova, Netherlands

  • Standard named IMGeo, information model geography
  • Integration of CityGML and IMGeo
  • IMGeo 2.0 established in 2012
  • OGC 3DIM adopted their approach
  • Port of Rotterdam uses in some cases physical models (!) to visualize underground 3D
  • Thanks to these efforts, stakeholders can now have "focused 3D ambitions"

Exploring cultural heritage resources in a 3D collaborative environment

Hidalgo, Spain

  • For presentation, analysis and exploration
  • Openwonderland, collaborative virtual world
  • They used "improvise" build-browse-behold java software to build and browse visualizations interactively

    • Demoed
  • Mostly at the conceptual stage - Interface not tested by users yet
  • Speaker not so comfortable in English

OpenBuildingModels - Towards a platform for crowdsourcing virtual 3D cities

Uden, university of Heidelberg

  • OpenStreetMap and the Google 3D Warehouse
  • OSM-3D: 3D globe based on OSM and SRTM-DEM data
  • Prerequisites for 3D VGI have never been this good: Kinect, Photosynth... Low cost hardware and service-based photogrammetry
  • Current issues:

    • No 3D support in OSM's simple data model
    • Not enough tools for users to contribute various 3D information in different LoDs
  • Vision for OpenBuildingModels

    • Free to use repository, with links to OSM
    • There is already multiple 3D models communities out there

      • 3D warehouse, opensceneryx.com, free3dmodels.org, shapeways.com, archive3d.net
    • Openbuildingmodels is a prototype at the moment
    • Using PostGIS, GeoServer, OpenLayers
    • Challenges: model georeference, and various other minor elements
  • Excellent presentation
  • Next big thing is 3D VGI

Mobile modeling for crowdsourcing building interior models

Rosser, uk

  • For exterior, Google Building Maker, for interior Trimble SketchUp (formerly Google SketchUp)
  • OpenRoomMap, extending OSM for indoor mapping - indoorosm
  • User control on sensitive data
  • Smartphones for data capture, but georeferencing remains an issue (no GPS indoor)

    • They built a prototype app for Android's galaxy nexus
  • Many challenges in the data capture, be to do many assumptions

Google Plus One