Stochastic Systems Group
Home Research Group Members Programs  
Demos Calendar Publications Mission Statement Alumni

SSG Seminar Abstract


Efficient Multi-Target Tracking Using Graphical Models

Zhexu (Michael) Chen
SSG, MIT


Multi-Target Tracking (MTT) has extensive military and surveillance applications. Despite being a well-studied problem, MTT remains very challenging. Although the Multi-Hypothesis Tracking (MHT) algorithm has been the preferred solution to MTT for the past 20 years, its computational complexity grows exponentially with the duration of the tracking window. It is also well known that conventional MHT has certain limitations, for instance, being unable to handle special situations, such as late data arrival and track stitching. We exploit the use of graphical models and message-passing algorithms to perform efficient MTT in linear time. Most importantly, our message-passing structure allows us to circumvent MHT's many inherent limitations.



Problems with this site should be emailed to jonesb@mit.edu