BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Dynamic Group Tracking - S. K. Pang
DTSTART:20080512T130000Z
DTEND:20080512T140000Z
UID:TALK11738@talks.cam.ac.uk
CONTACT:Taylan Cemgil
DESCRIPTION:For many surveillance applications\, targets of interest tend 
 to travel in groups\; for example\, a group of aircraft moving in a tight 
 formation or a convoy of vehicles moving along a road. There are two princ
 ipal reasons why we wish to model explicitly the behaviour of groups\, as 
 opposed to treating all\nobjects independently as in most multiple target 
 tracking approaches. The first is that the joint tracking of (a priori) de
 pendent objects within a group will lead to greater detection and\ntrackin
 g ability in hostile environments with high noise and low detection probab
 ilities. For example\, if we know several targets are in a group formation
 \, then we can infer some information of the positions and speed of those 
 targets with missing measurements due\nto poor detection probability\, giv
 en those target measurements that we do have. Similarly\, if a newly detec
 ted target appears close to an existing group\, the target can be initiali
 zed using the group velocity. Secondly\, we may hope to be able to infer a
 utomatically which objects are moving in formation as a group\, and which 
 are not. Thus we may be able to group objects according to intentionality\
 , behaviour and friend or foe\, all highly desirable inference outcomes fo
 r the modern tracking application.\n\nIn this talk\, we will define the pr
 oblem of dynamic group tracking. We will introduce the group dynamical mod
 els from a continuous time setting\, the interaction models for closely sp
 aced targets as well as a group structure transition model. We\nwill simul
 taneously tackle the problem of group structure inference and joint detect
 ion and tracking for group and individual targets within a Bayesian filter
 ing framework. We will present some tracking results based on simulations 
 as well as data from an actual GMTI radar. 
LOCATION:LR5\, Engineering\, Department of
END:VEVENT
END:VCALENDAR