Semantic retrieval of events from indoor surveillance video databases
Pattern Recognition Letters
Hi-index | 0.00 |
This paper presents a framework of a system for thequery and retrieval of video data based on video eventsin huge video repositories. The events are formulated usingdomain-independent event primitivies which are representedby spatio-temporal relationships between objects inthe video scenes. Complex events are expressible as combinationsof simpler events. This facilitates support of eventqueries from a variety of points of view. In addition, theframework is expected to provide adaptability of the frameworkto multiple domains.