STArT THINKING SEMINAR - Statistical aspects of computer vision: The problem of tracking in thermal video

Panagiotis Tsiamyrtzis

Assistant Professor

Dept. of Statistics

Athens University of Economics and Business

pt@aueb.gr

Abstract

In the area of computer vision one can come across of very interesting but quite challenging statistical problems. In this talk we will present the problem of tracking in videos.

No prior knowledge will be assumed. We will start by describing the problem and providing suggestions from the existing literature. Next, we will propose a novel template based tracking method that uses a network of independent particle filter trackers whose interactions are modeled using coalitional game theory. The proposed methodology updates the template, allowing the tracker to maintain pixel level accuracy even for long videos. Furthermore, it can negotiate surface deformations and occlusions.

The developed methodology can be applied to either visual or thermal videos and has been employed successfully to several real life problems. We will present shortly some of these applications in: lie detection, measuring vital signs from a distance, sleep studies, face recognition and measuring stress.

All this work has been developed in cooperation with several members of the Computational Physiology Lab (CPL) at the Computer Science Department of the University of Houston. 

Date:  17-5-2011 19:30-21:00

Location:  Classroom A32 Athens University of Economics and Business