Multiview Reconstruction by Gradual Enforcement of Relative Pose Transitivity

Result code in IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21230%2F08%3A03150826" target="_blank" >RIV/68407700:21230/08:03150826 - isvavai.cz</a>

Result on the web

—

DOI - Digital Object Identifier

—

Result language

angličtina

Original language name

Multiview Reconstruction by Gradual Enforcement of Relative Pose Transitivity

Original language description

It is known that the relative camera pose estimation is biased, e.g., due to noise in image correspondences. This bias prevents pairwise reconstructions from accurate alignment when estimating the multiview reconstruction. This work proposes a new methodfor making the relative camera poses more consistent. New constraints on transitivity of relative rotations, translations and their combination are proposed. An optimization based on the bundle adjustment starts with relative poses consistent with pairwise image measurements. The transitivity constraints are enforced gradually by enlarging their weights. Since the enlarging is done gradually, the relative poses are kept consistent with the data during the whole optimization, preventing so falling intodistant local minima. The technique is demonstrated on difficult wide base-line image sets to provide lower residuals and more accurate reconstructions.

Czech name

Multiview Reconstruction by Gradual Enforcement of Relative Pose Transitivity

Czech description

It is known that the relative camera pose estimation is biased, e.g., due to noise in image correspondences. This bias prevents pairwise reconstructions from accurate alignment when estimating the multiview reconstruction. This work proposes a new methodfor making the relative camera poses more consistent. New constraints on transitivity of relative rotations, translations and their combination are proposed. An optimization based on the bundle adjustment starts with relative poses consistent with pairwise image measurements. The transitivity constraints are enforced gradually by enlarging their weights. Since the enlarging is done gradually, the relative poses are kept consistent with the data during the whole optimization, preventing so falling intodistant local minima. The technique is demonstrated on difficult wide base-line image sets to provide lower residuals and more accurate reconstructions.

Type

O - Miscellaneous

CEP classification

JD - Use of computers, robotics and its application

OECD FORD branch

—

Project

<a href="/en/project/1ET101210406" target="_blank" >1ET101210406: Automatic 3D Virtual Model Builder from Photographs</a><br>

Continuities

R - Projekt Ramcoveho programu EK

Publication year

2008

Confidentiality

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů