Bundle Adjustment using Conjugate Gradients with Multiscale Preconditioning
Byröd, Martin LU and Åström, Karl LU
(2009)
British Machine Vision Conference, 2009
- Abstract
- Bundle adjustment is a key component of almost any feature based 3D reconstruction
system, used to compute accurate estimates of calibration parameters and structure and
motion configurations. These problems tend to be very large, often involving thousands
of variables. Thus, efficient optimization methods are crucial. The traditional Levenberg
Marquardt algorithm with a direct sparse solver can be efficiently adapted to the special
structure of the problem and works well for small to medium size setups. However, for
larger scale configurations the cubic computational complexity makes this approach pro-
hibitively expensive. The natural step here is to turn to iterative methods for... (More) - Bundle adjustment is a key component of almost any feature based 3D reconstruction
system, used to compute accurate estimates of calibration parameters and structure and
motion configurations. These problems tend to be very large, often involving thousands
of variables. Thus, efficient optimization methods are crucial. The traditional Levenberg
Marquardt algorithm with a direct sparse solver can be efficiently adapted to the special
structure of the problem and works well for small to medium size setups. However, for
larger scale configurations the cubic computational complexity makes this approach pro-
hibitively expensive. The natural step here is to turn to iterative methods for solving the
normal equations such as conjugate gradients. So far, there has been little progress in this
direction. This is probably due to the lack of suitable pre-conditioners, which are con-
sidered essential for the success of any iterative linear solver. In this paper, we show how
multi scale representations, derived from the underlying geometric layout of the problem,
can be used to dramatically increase the power of straight forward preconditioners such
as Gauss-Seidel. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/1612231
- author
- Byröd, Martin
LU
and Åström, Karl
LU
- organization
- publishing date
- 2009
- type
- Chapter in Book/Report/Conference proceeding
- publication status
- published
- subject
- keywords
- Computer vision, non-linear least squares problems, simultaneous localization and mapping, bundle adjustment
- host publication
- British Machine Vision Conference
- pages
- 10 pages
- conference name
- British Machine Vision Conference, 2009
- conference location
- London, United Kingdom
- conference dates
- 2009-09-07 - 2009-09-10
- external identifiers
-
- scopus:84898925190
- language
- English
- LU publication?
- yes
- id
- 16c6e967-f1b0-4461-bd6e-f502dae55a98 (old id 1612231)
- alternative location
- http://www.bmva.org/bmvc/2009/index.htm
- date added to LUP
- 2016-04-04 13:06:19
- date last changed
- 2022-04-24 02:33:17
@inproceedings{16c6e967-f1b0-4461-bd6e-f502dae55a98,
abstract = {{Bundle adjustment is a key component of almost any feature based 3D reconstruction<br/><br>
system, used to compute accurate estimates of calibration parameters and structure and<br/><br>
motion configurations. These problems tend to be very large, often involving thousands<br/><br>
of variables. Thus, efficient optimization methods are crucial. The traditional Levenberg<br/><br>
Marquardt algorithm with a direct sparse solver can be efficiently adapted to the special<br/><br>
structure of the problem and works well for small to medium size setups. However, for<br/><br>
larger scale configurations the cubic computational complexity makes this approach pro-<br/><br>
hibitively expensive. The natural step here is to turn to iterative methods for solving the<br/><br>
normal equations such as conjugate gradients. So far, there has been little progress in this<br/><br>
direction. This is probably due to the lack of suitable pre-conditioners, which are con-<br/><br>
sidered essential for the success of any iterative linear solver. In this paper, we show how<br/><br>
multi scale representations, derived from the underlying geometric layout of the problem,<br/><br>
can be used to dramatically increase the power of straight forward preconditioners such<br/><br>
as Gauss-Seidel.}},
author = {{Byröd, Martin and Åström, Karl}},
booktitle = {{British Machine Vision Conference}},
keywords = {{Computer vision; non-linear least squares problems; simultaneous localization and mapping; bundle adjustment}},
language = {{eng}},
title = {{Bundle Adjustment using Conjugate Gradients with Multiscale Preconditioning}},
url = {{https://lup.lub.lu.se/search/files/6053078/1612241.pdf}},
year = {{2009}},
}