Thesis Type: Doctorate
Institution Of The Thesis: Orta Doğu Teknik Üniversitesi, Faculty of Engineering, Department of Computer Engineering, Turkey
Approval Date: 2009
Student: YAKUP YILDIRIM
Consultant: ADNAN YAZICIAbstract:
Recent increase in the use of video in many applications has revealed the need for extracting the content in videos. Raw data and low-level features alone are not sufficient to fulfill the user's need; that is, a deeper understanding of the content at the semantic level is required. Currently, manual techniques are being used to bridge the gap between low-level representative features and high-level semantic content, which are inefficient, subjective and costly in time and have limitations on querying capabilities. Therefore, there is an urgent need for automatic semantic content extraction from videos. As a result of this requirement, we propose an automatic semantic content extraction system for videos in terms of object, event and concept extraction. We introduce a general purpose ontology-based video semantic content model that uses object definitions, spatial relations and temporal relations in event and concept definitions. Various relation types are defined to describe fuzzy spatio-temporal relations between ontology classes. Thus, the video semantic content model is utilized to construct domain ontologies. In addition, domain ontologies are enriched with rule definitions to lower spatial relation computation cost and to be able to define some complex situations more effectively. As a case study, we have performed a number experiments for event and concept extraction in videos for basketball and surveillance domains. We have obtained satisfactory precision and recall rates for object, event and concept extraction. A domain independent application for the proposed framework has been fully implemented and tested.