An Ontology for Depiction of Dynamic Scenes in Video
Zhuojun Li · TSpace (University of Toronto) · 2019
Smart sensor data in various domains are being generated and used for analysis. Motivated by the representation and understanding of video data generated from cameras, this work focuses on designing a first-order ontology that supports the interpretation of video depicting scenes with moving objects. We first revisit the Vision CardWorld Kernel Theory to identify and verify the formal incidence structures for scene, image, and depiction relations. Then, we integrate the kernel theory with the Process Specification Language (PSL) ontology to characterize change in the physical world. We also incorporate a location ontology with the kernel theory and introduce the formal axiomatization in mereotopology. Finally, we propose the Ontology for Video (OVid), which supports event recognition, by specifying the domain state ontology for the integration of the CardWorld Kernel Theory, Theory of Occlusion, Multidimensional Mereotopology, and Multidimensional Occupy Theory.