Toward Robust Online Visual Tracking

We pursue a research direction that will empower machines with simultaneous tracking and recognition capabilities similar to human cognition. Toward that, we develop algorithms that leverage prior knowledge/model obtained offline with information available online via novel learning algorithms. While humans can effortlessly locate moving objects in different environments, visual tracking remains one of the most important and challenging problems in computer vision. Robust cognitive visual tracking algorithms facilitate answering important questions regarding how objects move and interact in complex environments. They have broad applications including surveillance, navigation, human computer interfaces, object recognition, motion analysis and video indexing, to name a few.