Please use this identifier to cite or link to this item: https://doi.org/10.5194/isprs-annals-IV-2-W1-23-2016
DC FieldValue
dc.titleTOWARDS AUTOMATIC SEMANTIC LABELLING of 3D CITY MODELS
dc.contributor.authorRook M.
dc.contributor.authorBiljecki F.
dc.contributor.authorDiakité A.A.
dc.date.accessioned2018-10-05T08:05:46Z
dc.date.available2018-10-05T08:05:46Z
dc.date.issued2016
dc.identifier.citationRook M., Biljecki F., Diakité A.A. (2016). TOWARDS AUTOMATIC SEMANTIC LABELLING of 3D CITY MODELS. ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences 4 (2W1) : 23-30. ScholarBank@NUS Repository. https://doi.org/10.5194/isprs-annals-IV-2-W1-23-2016
dc.identifier.issn21949042
dc.identifier.urihttp://scholarbank.nus.edu.sg/handle/10635/148035
dc.description.abstractThe lack of semantic information in many 3D city models is a considerable limiting factor in their use, as a lot of applications rely on semantics. Such information is not always available, since it is not collected at all times, it might be lost due to data transformation, or its lack may be caused by non-interoperability in data integration from other sources. This research is a first step in creating an automatic workflow that semantically labels plain 3D city model represented by a soup of polygons, with semantic and thematic information, as defined in the CityGML standard. The first step involves the reconstruction of the topology, which is used in a region growing algorithm that clusters upward facing adjacent triangles. Heuristic rules, embedded in a decision tree, are used to compute a likeliness score for these regions that either represent the ground (terrain) or a RoofSurface. Regions with a high likeliness score, to one of the two classes, are used to create a decision space, which is used in a support vector machine (SVM). Next, topological relations are utilised to select seeds that function as a start in a region growing algorithm, to create regions of triangles of other semantic classes. The topological relationships of the regions are used in the aggregation of the thematic building features. Finally, the level of detail is detected to generate the correct output in CityGML. The results show an accuracy between 85 % and 99 % in the automatic semantic labelling on four different test datasets. The paper is concluded by indicating problems and difficulties implying the next steps in the research.
dc.publisherCopernicus GmbH
dc.sourceScopus
dc.typeConference Paper
dc.contributor.departmentDEPT OF ARCHITECTURE
dc.description.doi10.5194/isprs-annals-IV-2-W1-23-2016
dc.description.sourcetitleISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
dc.description.volume4
dc.description.issue2W1
dc.description.page23-30
Appears in Collections:Elements
Staff Publications

Show simple item record
Files in This Item:
File Description SizeFormatAccess SettingsVersion 
2016_3dgeoinfo_semantic_classification.pdf2.03 MBAdobe PDF

OPEN

NoneView/Download

SCOPUSTM   
Citations

3
checked on Sep 17, 2019

Page view(s)

62
checked on Sep 19, 2019

Download(s)

16
checked on Sep 19, 2019

Google ScholarTM

Check

Altmetric


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.