March 13, 2000, 1:00 PM
Toni Piirainen

University of Oulu,
Oulu, Finland

Video Indexing System

ABSTRACT

The motivation of this work is to get fully automated Video Indexing Systems that analyses and annotates the content of video sequences automatically without an operator. The basis for the system is done with supervised neural networks - Self Organizing Maps (SOM's). Video indexing makes the video sequences searchable. Other sytems like the ISLIP (www.islip.com) needs the operator to annotate the content of a video sequence. A basic presentation of the Self-Organizing MAP (SOM) will be demonstrated with a multidimensional vector data as in a two dimensional map. We use SOMs both as a classifier based only on sequence features, and as tree classifier, to distinguish between more general classes.