Modeling and Comparing Spatiotemporal Events
Current query systems for video databases rely heavily on structured schemas and often require user annotation of data. These systems need to be made more flexible and accessible to the diverse organizations that utilize them. By building on current work with lifelines, and incorporating new structures for organizing metadata, we can make great strides towards accomplishing this goal. This paper examines current methods for working with video queries, and explores how the use of hierarchical organization can improve on these methods. We look at how data cubes can be combined with image pyramids and scale space theory to facilitate this hierarchy. Finally, metadata structures and intelligent feedback systems are discussed and their uses both before and after the query are explained in the context of our system. This system represents a new way of looking at the information contained within a video sequence and the resultant ways in which the information should be organized.