Authors:
Xin Zhang
1
;
Tim Tutenel
2
;
Rong Mo
1
;
Rafael Bidarra
2
and
Willem F. Bronsvoort
2
Affiliations:
1
Northwestern Polytechnical University, China
;
2
Delft University of Technology, Netherlands
Keyword(s):
Model Classification, Segmentation, Annotation, Semantics.
Related
Ontology
Subjects/Areas/Topics:
Augmented, Mixed and Virtual Environments
;
Computer Vision, Visualization and Computer Graphics
;
Geometric Computing
;
Geometry and Modeling
;
Interactive Environments
;
Modeling and Algorithms
Abstract:
Semantics of 3D models is playing a crucial role in games and simulations. In this paper we propose a framework to specify semantics of large sets of 3D models with minimal human involvement. The framework consists of three modules: classification, segmentation and annotation. We associate a few models with tags representing their classes and classify the other models automatically. Once all models have been classified in different groups, we take a certain number of models as template models in each group, and segment these template models interactively. We then use the segmentation method (and parameters) of the template models to segment the rest of the models of the same group automatically. We annotate the interactively segmented parts and use an attributed adjacency graph to represent them. Automatic annotation of the rest of the models is then performed by subgraph matching. Experiments show that the proposed framework can effectively specify semantics of large sets of 3D mode
ls.
(More)