SEMANTIC ALIGNMENT OF DOCUMENTS WITH 3D CITY MODELS Camille Tardy 1, Laurent Moccozet 2 and Gilles Falquet 1 University of Geneva 1 Institute ICLE 2 Institute of Services Science FACULTÉ DES SCIENCES ÉCONOMIQUES ET SOCIALES Institut de recherche en ingénierie des connaissances et logiques de l espace
A Digital Library of Urban resources Documents 3D city Models http://www.lviz.org/wp- content/uploads/2009/03/vancouver1980.png 2
Context Requires access to Task http://www.lviz.org/wp- content/uploads/2009/03/vancouver1980.png 3
Problem Statement Navigates/ Looks at Establish semantic links between 3DCM objects and documents http://www.lviz.org/wp- content/uploads/2009/03/vancouver1980.png 4
ContextualisaFon Represent the documents in context Visually link documents to the 3D object Requires : 1. Compute the semantic alignment relationships 2. Compute the user interface layout 5
Related Work 3D visualisation : NIRVE [1], Periscope[2] Contextualisation in space : Geocoding process in GIS [3] Visualisation : World Explorer [4] 6
A Model to Compute the Links Coverage Geonames CityGML Urbamet Annotation Domain Ontologies Link document- object http://www.lviz.org/wp- content/uploads/2009/03/vancouver1980.png Semantic alignment of documents with 3D city models, C. Tardy et al. 7
IndexaFon and AnnotaFon Documents are indexed Annotation RDF \ile Coverage Space Time 3d object annotation CityGML representation and Geonames identi\ication 8
Ontology Alignment Indexation vocabularies CityGML Geonames Urbamet Domain Ontologies 9
IdenFficaFon Geonames CityGML Urbamet Domain Ontologies Link document- object http://www.lviz.org/wp- content/uploads/2009/03/vancouver1980.png 10
IdenFfying city objects From CityGML to Geonames Position Function Rules to solve ambiguities 11
IdenFficaFon of City Objects 12
A Model to Compute the Links Coverage Geonames CityGML Urbamet Annotation Domain Ontologies Link document- object http://www.lviz.org/wp- content/uploads/2009/03/vancouver1980.png Semantic alignment of documents with 3D city models, C. Tardy et al. 13
Find the connecfons Type of links Explicit (Boolean value) Direct or indirect connection Suggestion (Weight value) Through non- geographic entities Object is not contained in the 3DCM 14
Alignment Algorithm Input Document corpora (indexed) 3D city model (indexed) Output Matrix of links M(doc,obj) Focus on the spatial coverage 15
Algorithm One document, one object Cov(obj) Cov(doc) or vice versa CityGML/Geonames Search_link function 4 cases : 1. Direct link 2. Direct link by class 3. Link through non- geographic entities 4. Link by spatial proximity 16
Direct Link Cov(obj) G : Geneva airport G: Geneva airport Cov(doc) M(doc,obj)= true 17
Direct Link by Class Cov(obj) G : Geneva airport airport_ building M(doc,obj)=true Cov(doc) 18
Link through Non- geographic EnFFes Cov(obj) G: HUG surgery Idx(doc) M(doc,obj)=1/semantic_distance( surgery, HUG ) 19
Link by SpaFal Proximity e not in cov(model) e and f are spatially close G: e G: e Cov(doc) Cov(model) = f (geoname) M(doc,obj)= 1/euclidean_distance(e,f) http://i149.photobucket.com/albums/s62/workingfilms/aucklandcity2.jpg 20
ParFcular SituaFons The object is a group Storeys 21
Example : Geneva Airport http://www.journal- aviation.com/ Cov(doc) = Switzerland (geoname) Idx(doc) = Swiss, airline, tablets, crew 22
Example : Geneva Airport Cov(obj) = BuildingFunctionType : Airport_Building (citygml) Cov(model) = Geneva Canton (geoname) e (geoname ) = label : Aéroport Genève- Cointrin, feature code : S.AIRP (Airport) 23
Example : Geneva Airport Switzerland contains Geneva Canton Idx(doc) contains airline Link between doc and all airports objects in the model M(doc,obj)=true - > explicit_link 24
Example : University of Geneva www.unige.ch Cov(doc) = Uni Dufour (geoname), Storey : name = 1 (CityGML) 25
Example : University of Geneva Obj is a group of 4 storeys : v o4 Storey : name = 3 v o3 Storey : name = 2 v o2 Storey : name = 1 v o1 Storey : name = 0 Cov(obj) = Uni Dufour (geoname) http://sketchup.google.com/3dwarehouse/details?mid=da6dc\b704f4294c9202d3aef1555f0c 26
Example : University of Geneva Cov(doc) and Cov(obj) contains Uni Dufour Cov(doc) - > Storey Search link with obj sub- objects Cov(o2)= Uni Dufour, Storey : name = 1 M(doc,o2)=true 27
Conclusion Semantic indexing Identify links through numerous connections Future Work : Evaluation via a use case Deal with fuzzy area Develop the different visualisation technics Develop the query system 28
References 1. W. Wiza, K. Walczak, W. Cellary, Proc. of the 9th Int. Conf. on 3D Web Technology. Periscope - A System for Adaptive 3D Visualization of Search Results. 29 40. (2004) 2. J. Cugini, S. Laskowski, M. Sebrechts, Proc. of IST/SPIE's 12th Annual Int. Symposium: Electronic Imaging. Design of 3D visualization of search results: Evolution and evaluation. 23 28. (2000) 3. C.B. Jones, A.I. Abdelmoty, D. Finch, G. Fu, S. Vaid, Proc. of the 3rd Int. Conf. on Geographic Information Science. The SPIRIT Spatial Search Engine: Architecture, Ontologies and Spatial Indexing. 125 139. (2004) 4. S. Ahern, M. Naaman, R. Nair, J.H.I. Yang, Proc. of the 7th JCDL. World explorer: visualizing aggregate data from unstructured text in geo- referenced collections. 1 10. (2007) 29