Visual contact estimation for navigation and perception

Chris McCarthy (The Australian National University)

DATE: 2009-09-17
TIME: 16:00:00 - 17:00:00
LOCATION: NICTA - 7 London Circuit

ABSTRACT:
A fundamental capability of any visual navigation system is the perception of potential contact with surfaces in the environment. Ecological studies of animal behaviour have highlighted the importance of motion perception in achieving this. The visual measure of motion, optical flow, provides useful cues that may be directly used for perceiving scene structure, and for visually guiding self-motion. In this talk I will overview the major contributions of my (soon to be submitted) PhD thesis, which proposes new biologically-inspired strategies for estimating surface contact from optical flow. I focus primarily on the estimation of time-to-contact for vision-guided docking and landing. I present a robust strategy for docking a mobile robot with an upright surface using optical flow field divergence. I then present a unified strategy for docking and landing with surfaces of arbitrary orientation using flow field divergence under a spherical projection model. I show that spherical projection offers distinct advantages over perspective projection when estimating time-to-contact from flow divergence. I also report on work exploring the use of a spherical projection model for real-time, 3D depth map recovery from optical flow, and discuss preliminary work investigating the use of optical flow to predict the time and location of impact of an incoming, self-moving objects. This work is inspired by the observed behaviour of bi-modal neurons in the pre-motor cortex of primates.

BIO:
Chris McCarthy received the B.Sc degree with honours in 1999, and a Master of Computer Science degree in 2005 from the University of Melbourne. In 2000 he worked as an International Fellow in the Centre for Computer Studies, at Ngee Ann Polytechnic, Singapore. From 2001-2005, he was a Senior Tutor and Associate Lecturer with the Department of Computer Science and Software Engineering at The University of Melbourne. Since 2005, he has been working towards a PhD degree with the Australian National University and NICTA's Canberra Research Laboratory. This time has included a 6 month placement in the Robotics, Brain and Cognitive Sciences lab at the Italian Institute of Technology, Genova in 2007. In April 2009 he took up the position of Researcher with NICTA's Visual Processing for the Bionic Eye (VIBE) project. His research interests are in vision for robot navigation and human assistance, and computational models of biological vision.

	Skip Navigation CECS Home \| ANU Home \| Search ANU \| HORUS \| Staff Home
	Seminars @ CECS ANU College of Engineering and Computer Science