期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2014
卷号：64
期号：2
出版社：Journal of Theoretical and Applied
摘要：As a large amount of video based applications becomes widely available, there is a need to extract the semantic data in videos. Manual techniques, which are inefficient and subjective and costly in time, bound querying capabilities and bridge gap between low-level representative feature and high level semantic content. For the deeper understanding apart from the raw data and low-level features, the content at semantic level is required. The existing periodicity mining was based on audio retrieval which cannot be applied to the video. Optical flow algorithm is used to track the direction of the spatial movements between two image frames. Ontology provides a vast domain applicable rule production set that allows the user to construct the domain. Rule based model use the spatial/temporal relations to extract the definitions. The moving regions are extracted by the Graph-Based Visual Saliency (GPVS) algorithm and it detects the temporal locations. The phase correlation method is a well-known image mining technique in multimedia data which has main application in coordination of frames.
关键词：Spatial Movement; Ontology; Spatial/Temporal Relation; Rule Based Model; Phase Correlation.
