Associate Professor of Computer Science
The central theme of Duric's computer vision research is understanding the motions of humans and of vehicles driven by humans. Application areas for this research include smart rooms, video surveillance and monitoring, human-computer interaction, secure driving/ intelligent highways, and video coding.
In his work he has been applying techniques from such domains as theoretical kinematics and dance notation to analyze physical and geometrical constraints on the motions of humans and vehicles. This research will result in methods that significantly improve on currently available techniques for computing human-generated motions of objects.
In his previous work he developed the Frenet-Serret and Darboux motion models to describe physically possible motions of tools and vehicles. In his current workhe is extending this research to understanding human motions. In particular Duric is investigating how dance notation can be used to describe human motions in such domains as gestures and sports.
In the domain of human motion understanding Duric is interested in gestures and simple activities performed by small numbers of humans. In the gesture domain a single human is viewed by one or more cameras. The human uses upper body gestures such as posture, head pose (nods, etc.), shoulder movements (shrugs, etc.), hand and arm movements, palm facings, finger pointings, and so on to convey a message. In the domain of "general'' human motions and sports Duric is interested in simple movements that are performed in an approximately upright position, such as walking, running, etc.
Media Contact: Preston Williams, 703-993-9376, email@example.com