This page is provided for historical and archival purposes only. While the seminar dates are correct, we offer no guarantee of informational accuracy or link validity. Contact information for the speakers, hosts and seminar committee are certainly out of date.
Dan Huttenlocher Computer Science Department Cornell University and Xerox Palo Alto Research Center
We are investigating the extent to which two-dimensional representations are applicable to tasks that involve interacting with the three-dimensional world. We have been using two-dimensional geometry extracted from images to solve a number of problems, including tracking moving objects, segmenting images based on motion, and navigating a mobile robot to a target. Our methods are based on comparing portions of one image with another, under various transformations, using the generalized Hausdorff measure. This measure determines the resemblance of point sets (binary images) by examining the fraction of points in one set that are near points in the other (and perhaps vice versa). Unlike most measures used in model-based matching, this measure is not based on computing a correspondence between points in the two sets.
In this talk, I will describe the generalized Hausdorff measure, discuss how to compute it efficiently, and then present applications of the measure to motion tracking and visually guided navigation problems.
Host: Yangsheng Xu (xu+@cs)/Martial Hebert (hebert@cs) Appointment: Ava Cruse (avac@cs)/Marie Elm (mke@cs)