Globally Consistent Dense Real-Time 3D Reconstruction from RGBD Data

Rafael Jakob Weilharter; Fabian Schenk; Friedrich Fraundorfer

doi:10.3217/978-3-85125-603-1-25

Globally Consistent Dense Real-Time 3D Reconstruction from RGBD Data

Rafael Jakob Weilharter, Fabian Schenk, Friedrich Fraundorfer

Institute of Computer Graphics and Vision (7100)

Research output: Chapter in Book/Report/Conference proceeding › Conference paper › peer-review

Abstract

In this work, we present a dense 3D reconstruction
framework for RGBD data that can handle loop closure and
pose updates online. Handling updates online is essential to get
a globally consistent 3D reconstruction in real-time. We also
introduce fused depth maps for each keyframe that contain
the fused depths of all associated frames to greatly increase
the speed for model updates. Furthermore, we show how we
can use integration and de-integration in a volumetric fusion
system to adjust our model to online updated camera poses.
We build our system on top of the InfiniTAM framework to
generate a model from the semi-dense, keyframe based ORB
SLAM2. We extensively evaluate our system on real world and
synthetic generated RGBD data regarding tracking accuracy
and surface reconstruction.

Original language	English
Title of host publication	Proceedings of the OAGM Workshop 2018
Subtitle of host publication	Medical Image Analysis
Editors	Martin Welk, Martin Urschler, Peter M. Roth
Place of Publication	Graz
Publisher	Verlag der Technischen Universität Graz
Pages	121-127
Number of pages	7
ISBN (Electronic)	978-3-85125-603-1
DOIs	https://doi.org/10.3217/978-3-85125-603-1-25
Publication status	Published - 2018
Event	42nd Annual Workshop of the Austrian Association for Pattern Recognition: Medical Image Analysis: ÖAGM 2018 - Hall/Tirol, Austria Duration: 15 May 2018 → 16 May 2018

Workshop

Workshop	42nd Annual Workshop of the Austrian Association for Pattern Recognition: Medical Image Analysis
Country/Territory	Austria
City	Hall/Tirol
Period	15/05/18 → 16/05/18

Keywords

3D reconstruction
RGBD
volumetric fusion
SLAM

Access to Document

10.3217/978-3-85125-603-1-25Licence: CC BY 4.0

CSISmartScan - Integrated 3D scene and documentation
Fraundorfer, F.
1/10/15 → 30/09/17
Project: Research project

OCG Best Paper Award
Weilharter, Rafael Jakob (Recipient), Schenk, Fabian (Recipient) & Fraundorfer, Friedrich (Recipient), 2018
Prize: Prizes / Medals / Awards

Cite this

Globally Consistent Dense Real-Time 3D Reconstruction from RGBD Data. / Weilharter, Rafael Jakob; Schenk, Fabian; Fraundorfer, Friedrich.
Proceedings of the OAGM Workshop 2018: Medical Image Analysis. ed. / Martin Welk; Martin Urschler; Peter M. Roth. Graz: Verlag der Technischen Universität Graz, 2018. p. 121-127.

Research output: Chapter in Book/Report/Conference proceeding › Conference paper › peer-review

Weilharter, RJ, Schenk, F & Fraundorfer, F 2018, Globally Consistent Dense Real-Time 3D Reconstruction from RGBD Data. in M Welk, M Urschler & PM Roth (eds), Proceedings of the OAGM Workshop 2018: Medical Image Analysis. Verlag der Technischen Universität Graz, Graz, pp. 121-127, 42nd Annual Workshop of the Austrian Association for Pattern Recognition: Medical Image Analysis, Hall/Tirol, Austria, 15/05/18. https://doi.org/10.3217/978-3-85125-603-1-25

@inproceedings{0b43acf3f4f945bab736b78ea7e2e0dc,

title = "Globally Consistent Dense Real-Time 3D Reconstruction from RGBD Data",

abstract = "In this work, we present a dense 3D reconstructionframework for RGBD data that can handle loop closure andpose updates online. Handling updates online is essential to geta globally consistent 3D reconstruction in real-time. We alsointroduce fused depth maps for each keyframe that containthe fused depths of all associated frames to greatly increasethe speed for model updates. Furthermore, we show how wecan use integration and de-integration in a volumetric fusionsystem to adjust our model to online updated camera poses.We build our system on top of the InfiniTAM framework togenerate a model from the semi-dense, keyframe based ORBSLAM2. We extensively evaluate our system on real world andsynthetic generated RGBD data regarding tracking accuracyand surface reconstruction.",

keywords = "3D reconstruction, RGBD, volumetric fusion, SLAM",

author = "Weilharter, {Rafael Jakob} and Fabian Schenk and Friedrich Fraundorfer",

year = "2018",

doi = "10.3217/978-3-85125-603-1-25",

language = "English",

pages = "121--127",

editor = "Martin Welk and Martin Urschler and Roth, {Peter M.}",

booktitle = "Proceedings of the OAGM Workshop 2018",

publisher = "Verlag der Technischen Universit{\"a}t Graz",

note = "42nd Annual Workshop of the Austrian Association for Pattern Recognition: Medical Image Analysis : {\"O}AGM 2018 ; Conference date: 15-05-2018 Through 16-05-2018",

}

TY - GEN

T1 - Globally Consistent Dense Real-Time 3D Reconstruction from RGBD Data

AU - Weilharter, Rafael Jakob

AU - Schenk, Fabian

AU - Fraundorfer, Friedrich

PY - 2018

Y1 - 2018

N2 - In this work, we present a dense 3D reconstructionframework for RGBD data that can handle loop closure andpose updates online. Handling updates online is essential to geta globally consistent 3D reconstruction in real-time. We alsointroduce fused depth maps for each keyframe that containthe fused depths of all associated frames to greatly increasethe speed for model updates. Furthermore, we show how wecan use integration and de-integration in a volumetric fusionsystem to adjust our model to online updated camera poses.We build our system on top of the InfiniTAM framework togenerate a model from the semi-dense, keyframe based ORBSLAM2. We extensively evaluate our system on real world andsynthetic generated RGBD data regarding tracking accuracyand surface reconstruction.

AB - In this work, we present a dense 3D reconstructionframework for RGBD data that can handle loop closure andpose updates online. Handling updates online is essential to geta globally consistent 3D reconstruction in real-time. We alsointroduce fused depth maps for each keyframe that containthe fused depths of all associated frames to greatly increasethe speed for model updates. Furthermore, we show how wecan use integration and de-integration in a volumetric fusionsystem to adjust our model to online updated camera poses.We build our system on top of the InfiniTAM framework togenerate a model from the semi-dense, keyframe based ORBSLAM2. We extensively evaluate our system on real world andsynthetic generated RGBD data regarding tracking accuracyand surface reconstruction.

KW - 3D reconstruction

KW - RGBD

KW - volumetric fusion

KW - SLAM

U2 - 10.3217/978-3-85125-603-1-25

DO - 10.3217/978-3-85125-603-1-25

M3 - Conference paper

SP - 121

EP - 127

BT - Proceedings of the OAGM Workshop 2018

A2 - Welk, Martin

A2 - Urschler, Martin

A2 - Roth, Peter M.

PB - Verlag der Technischen Universität Graz

CY - Graz

T2 - 42nd Annual Workshop of the Austrian Association for Pattern Recognition: Medical Image Analysis

Y2 - 15 May 2018 through 16 May 2018

ER -

Globally Consistent Dense Real-Time 3D Reconstruction from RGBD Data

Abstract

Workshop

Keywords

Access to Document

Fingerprint

Projects

CSISmartScan - Integrated 3D scene and documentation

Prizes

OCG Best Paper Award

Cite this