Understanding Spatial Knowledge : An Ontology-Based Representation for Object Identification Understanding Spatial Knowledge : An Ontology-Based Representation for Object Identification (Special Issue on Computer Vision and Applications)
Access this Article
Search this Article
Spatial descriptions are one of the most effective methods to enable interlocutors to identify which object is being discussed in discourse. In this paper, we propose a framework that can identify an object whose positional relation with another object is indicated verbally by a human. To this end, we construct a spatial knowledge ontology. The ontology is enriched by Description Logic (DL) of concepts, which allows discovering hidden knowledge. We also propose a Spatial Object Dataset that is specifically tailored for our experiments with ontological structures. The dataset currently contains 130 objects and in total of 720 images for object recognition and 360 scenes for spatial recognition. Preliminary experimental results confirmed that the system was able to correctly recognizes human descriptions and identify unknown objects and that understanding human spatial descriptions is efficient for human-machine interaction.
- IIEEJ transactions on image electronics and visual computing
IIEEJ transactions on image electronics and visual computing 3(2), 150-163, 2015
The Institute of Image Electronics Engineers of Japan