Entropy-based visual servoing and tracking

Amaury Dame (INRIA Rennes, France)

COMPUTER VISION AND ROBOTICS SERIES

DATE: 2010-01-28
TIME: 16:00:00 - 17:00:00
LOCATION: NICTA - 7 London Circuit
CONTACT: JavaScript must be enabled to display this email address.

ABSTRACT:
In this paper we propose a unified approach based on mutual information that allows both differential tracking and visual servoing. Mutual information (MI) is widely used in multi-modal image registration (medical applications) since it is robust to large changes in the lighting condition and to a wide class of non-linear image transformations. In this seminar we present how mutual information can be used as a new visual feature for visual servoing that allows us to build a new control law to control the 6 degrees of freedom of a robot. Among various advantages regarding robustness, this approach does not require any matching nor tracking step. We also present how MI can be used to estimate parametric motion models to build a robust differential tracker.
BIO:
Amaury Dame works in the LAGADIC research team in INRIA Rennes (France) under the direction of Francois Chaumette. He joined the team as a PhD student in informatics under the supervision of Eric Marchand on the development of robust tracking algorithms for vehicle navigation using visual servoing approaches. Amaury received an Engineering Master degree in industrial computer science and electronics from INSA Rennes France (INSA) and a Research Master degree in signal teledetection and image processing (STI) from the University of Rennes 1.