Moving target indication (MTI) sensors provide surveillance
information about ground traffic.  We are interested in discovering
and explaining unusual traffic patterns observed over long periods of
time.

A priori knowledge about traffic patterns is limited and clutter
(background traffic) may have a complex structure.  We demonstrate how
clustering techniques can be used in order to identify different
classes of objects and understand interactions between them, and
discuss arising problems in search and generalization of associations
and feature selection.