The New Modeling for Semantic Representation of Moving Objects in Video

Abstract

There are many researches about video retrieval using spatio-temporal relations. Most of them studied simple trajectory or content-based retrieval but without semantic motion. Therefore, in this paper, we propose a new modeling for semantic representation of moving object using new topological relations, directional relations. Especially, we described directional relations based on TSR for measuring and storing similarity of object's direction. Also, we built hierarchical structure by combining motions with motion verbs to describe semantic relations of moving objects. In the future work, we'll build knowledge base such as motion ontology and apply it for automatic summarization or annotation system.