Browse Prior Art Database

Method for Determining the Velocity of a Mobile Robot by Using Dynamic Stereo Vision

IP.com Disclosure Number: IPCOM000109252D
Original Publication Date: 1992-Aug-01
Included in the Prior Art Database: 2005-Mar-23
Document File: 2 page(s) / 93K

Publishing Venue

IBM

Related People

Echigo, T: AUTHOR

Abstract

A program is disclosed for determining the velocity of a mobile robot from the three-dimensional motion of feature points whose correspondences are found by comparing time-varying images from two cameras mounted on the mobile robot.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Method for Determining the Velocity of a Mobile Robot by Using Dynamic Stereo Vision

       A program is disclosed for determining the velocity of a
mobile robot from the three-dimensional motion of feature points
whose correspondences are found by comparing time-varying images from
two cameras mounted on the mobile robot.

      The proposed method employs two cameras whose axes of lenses
are parallel.  When the mobile robot moves, time-varying image
sequences can be captured from each camera, as shown in Fig. 1.  The
problem is then to find correspondences of multiple image flows
between the two cameras.  Corresponding points (xl, yl) and (xr, yr)
in the left and right images at an arbitrary time tk meet the
epipolar constraint.  Accordingly, corresponding image flows that
include the above points must satisfy the epipolar constraint at all
sampling points, because the geometric relation of the two cameras is
constant even though the robot moves.  However, this constraint is a
sufficient but not a necessary condition for corresponding image
flows. flows.  Consequently, multiple candidates, including erroneous
ones, that meet the epipolar constraint may remain for finding
correspondences.

      In the next step, the 3D positions of all corresponding
candidates at sampling time tk are calculated.  They are represented
as a perfect graph in which the weight of edges between calculated
feature points are relative distances.  At sampling time tk+1 a
perfect graph can be obtained in the same way.  Fig. 2 illustrates
the relationship betwee...