Wide Area Camera Calibration Using Virtual Calibration Objects

Xing Chen, James Davis, Philipp Slusallek 


Abstract

This paper introduces a method to calibrate a wide area  system of
unsynchronized cameras with respect to a single  global coordinate
system.  The method is simple and does  not require the physical
construction of a large  calibration object. The user need only wave
an  identifiable point in front of all cameras. The method  generates a
rough estimate o f camera pose by first  performing pair -wise
structure -from-motion on  observed  points, and then combining the
pair -wise registrations  into a single coordinate frame. Using the
initial camera  pose, the moving point can be tracked in world space.
The  path of  the point defines a "virtual calibration object"  which
can be used to improve the initial estimates of  camera pose. Iterating
the above process yields a more  precise estimate of both camera pose
and the point path.  Experimental results show that it perform s as
well as  calibration from a physical target, in cases where all
cameras share some common working volume. We then  demonstrate its
effectiveness in wide area settings by  calibrating a system of cameras
having non -overlapping  fields of view, a situation  where traditional
methods  cannot be applied directly.

CVPR 2000 paper

[PDF 5.5 MB]
[PS 7.3 MB]
 

Poster from CVPR 2000

[PDF 4.9MB]