3D Fluid Flow Estimation with Integrated Particle Reconstruction

Katrin Lasinger, Christoph Vogel, Thomas Pock, Konrad Schindler

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

Abstract

The standard approach to densely reconstruct the motion in a volume of fluid is to inject high-contrast tracer particles and record their motion with multiple high-speed cameras. Almost all existing work processes the acquired multi-view video in two separate steps: first, a per-frame reconstruction of the particles, usually in the form of soft occupancy likelihoods in a voxel representation; followed by 3D motion estimation, with some form of dense matching between the precomputed voxel grids from different time steps. In this sequential procedure, the first step cannot use temporal consistency considerations to support the reconstruction, while the second step has no access to the original, high-resolution image data. We show, for the first time, how to jointly reconstruct both the individual tracer particles and a dense 3D fluid motion field from the image data, using an integrated energy minimization. Our hybrid Lagrangian/Eulerian model explicitly reconstructs individual particles, and at the same time recovers a dense 3D motion field in the entire domain. Making particles explicit greatly reduces the memory consumption and allows one to use the high-resolution input images for matching. Whereas the dense motion field makes it possible to include physical a-priori constraints and account for the incompressibility and viscosity of the fluid. The method exhibits greatly (≈70) improved results over a recent baseline with two separate steps for 3D reconstruction and motion estimation. Our results with only two time steps are comparable to those of state-of-the-art tracking-based methods that require much longer sequences.

Original languageEnglish
Title of host publicationPattern Recognition - 40th German Conference, GCPR 2018, Proceedings
EditorsMario Fritz, Andrés Bruhn, Thomas Brox
PublisherSpringer-Verlag Italia
Pages315-332
Number of pages18
ISBN (Print)9783030129385
DOIs
Publication statusPublished - 1 Jan 2019
Event40th German Conference on Pattern Recognition, GCPR 2018 - Stuttgart, Germany
Duration: 9 Oct 201812 Oct 2018

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11269 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference40th German Conference on Pattern Recognition, GCPR 2018
CountryGermany
CityStuttgart
Period9/10/1812/10/18

Fingerprint

Fluid Flow
Flow of fluids
Motion estimation
Fluids
Motion
Motion Estimation
Voxel
Fluid
High speed cameras
Image resolution
High Resolution
Sequential Procedure
High-speed Camera
Incompressibility
Viscosity
Energy Minimization
3D Reconstruction
Data storage equipment
Baseline
Likelihood

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Lasinger, K., Vogel, C., Pock, T., & Schindler, K. (2019). 3D Fluid Flow Estimation with Integrated Particle Reconstruction. In M. Fritz, A. Bruhn, & T. Brox (Eds.), Pattern Recognition - 40th German Conference, GCPR 2018, Proceedings (pp. 315-332). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11269 LNCS). Springer-Verlag Italia. https://doi.org/10.1007/978-3-030-12939-2_22

3D Fluid Flow Estimation with Integrated Particle Reconstruction. / Lasinger, Katrin; Vogel, Christoph; Pock, Thomas; Schindler, Konrad.

Pattern Recognition - 40th German Conference, GCPR 2018, Proceedings. ed. / Mario Fritz; Andrés Bruhn; Thomas Brox. Springer-Verlag Italia, 2019. p. 315-332 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11269 LNCS).

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

Lasinger, K, Vogel, C, Pock, T & Schindler, K 2019, 3D Fluid Flow Estimation with Integrated Particle Reconstruction. in M Fritz, A Bruhn & T Brox (eds), Pattern Recognition - 40th German Conference, GCPR 2018, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11269 LNCS, Springer-Verlag Italia, pp. 315-332, 40th German Conference on Pattern Recognition, GCPR 2018, Stuttgart, Germany, 9/10/18. https://doi.org/10.1007/978-3-030-12939-2_22
Lasinger K, Vogel C, Pock T, Schindler K. 3D Fluid Flow Estimation with Integrated Particle Reconstruction. In Fritz M, Bruhn A, Brox T, editors, Pattern Recognition - 40th German Conference, GCPR 2018, Proceedings. Springer-Verlag Italia. 2019. p. 315-332. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-030-12939-2_22
Lasinger, Katrin ; Vogel, Christoph ; Pock, Thomas ; Schindler, Konrad. / 3D Fluid Flow Estimation with Integrated Particle Reconstruction. Pattern Recognition - 40th German Conference, GCPR 2018, Proceedings. editor / Mario Fritz ; Andrés Bruhn ; Thomas Brox. Springer-Verlag Italia, 2019. pp. 315-332 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{212373654754496e8e6997be23097611,
title = "3D Fluid Flow Estimation with Integrated Particle Reconstruction",
abstract = "The standard approach to densely reconstruct the motion in a volume of fluid is to inject high-contrast tracer particles and record their motion with multiple high-speed cameras. Almost all existing work processes the acquired multi-view video in two separate steps: first, a per-frame reconstruction of the particles, usually in the form of soft occupancy likelihoods in a voxel representation; followed by 3D motion estimation, with some form of dense matching between the precomputed voxel grids from different time steps. In this sequential procedure, the first step cannot use temporal consistency considerations to support the reconstruction, while the second step has no access to the original, high-resolution image data. We show, for the first time, how to jointly reconstruct both the individual tracer particles and a dense 3D fluid motion field from the image data, using an integrated energy minimization. Our hybrid Lagrangian/Eulerian model explicitly reconstructs individual particles, and at the same time recovers a dense 3D motion field in the entire domain. Making particles explicit greatly reduces the memory consumption and allows one to use the high-resolution input images for matching. Whereas the dense motion field makes it possible to include physical a-priori constraints and account for the incompressibility and viscosity of the fluid. The method exhibits greatly (≈70) improved results over a recent baseline with two separate steps for 3D reconstruction and motion estimation. Our results with only two time steps are comparable to those of state-of-the-art tracking-based methods that require much longer sequences.",
author = "Katrin Lasinger and Christoph Vogel and Thomas Pock and Konrad Schindler",
year = "2019",
month = "1",
day = "1",
doi = "10.1007/978-3-030-12939-2_22",
language = "English",
isbn = "9783030129385",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer-Verlag Italia",
pages = "315--332",
editor = "Mario Fritz and Andr{\'e}s Bruhn and Thomas Brox",
booktitle = "Pattern Recognition - 40th German Conference, GCPR 2018, Proceedings",
address = "Italy",

}

TY - GEN

T1 - 3D Fluid Flow Estimation with Integrated Particle Reconstruction

AU - Lasinger, Katrin

AU - Vogel, Christoph

AU - Pock, Thomas

AU - Schindler, Konrad

PY - 2019/1/1

Y1 - 2019/1/1

N2 - The standard approach to densely reconstruct the motion in a volume of fluid is to inject high-contrast tracer particles and record their motion with multiple high-speed cameras. Almost all existing work processes the acquired multi-view video in two separate steps: first, a per-frame reconstruction of the particles, usually in the form of soft occupancy likelihoods in a voxel representation; followed by 3D motion estimation, with some form of dense matching between the precomputed voxel grids from different time steps. In this sequential procedure, the first step cannot use temporal consistency considerations to support the reconstruction, while the second step has no access to the original, high-resolution image data. We show, for the first time, how to jointly reconstruct both the individual tracer particles and a dense 3D fluid motion field from the image data, using an integrated energy minimization. Our hybrid Lagrangian/Eulerian model explicitly reconstructs individual particles, and at the same time recovers a dense 3D motion field in the entire domain. Making particles explicit greatly reduces the memory consumption and allows one to use the high-resolution input images for matching. Whereas the dense motion field makes it possible to include physical a-priori constraints and account for the incompressibility and viscosity of the fluid. The method exhibits greatly (≈70) improved results over a recent baseline with two separate steps for 3D reconstruction and motion estimation. Our results with only two time steps are comparable to those of state-of-the-art tracking-based methods that require much longer sequences.

AB - The standard approach to densely reconstruct the motion in a volume of fluid is to inject high-contrast tracer particles and record their motion with multiple high-speed cameras. Almost all existing work processes the acquired multi-view video in two separate steps: first, a per-frame reconstruction of the particles, usually in the form of soft occupancy likelihoods in a voxel representation; followed by 3D motion estimation, with some form of dense matching between the precomputed voxel grids from different time steps. In this sequential procedure, the first step cannot use temporal consistency considerations to support the reconstruction, while the second step has no access to the original, high-resolution image data. We show, for the first time, how to jointly reconstruct both the individual tracer particles and a dense 3D fluid motion field from the image data, using an integrated energy minimization. Our hybrid Lagrangian/Eulerian model explicitly reconstructs individual particles, and at the same time recovers a dense 3D motion field in the entire domain. Making particles explicit greatly reduces the memory consumption and allows one to use the high-resolution input images for matching. Whereas the dense motion field makes it possible to include physical a-priori constraints and account for the incompressibility and viscosity of the fluid. The method exhibits greatly (≈70) improved results over a recent baseline with two separate steps for 3D reconstruction and motion estimation. Our results with only two time steps are comparable to those of state-of-the-art tracking-based methods that require much longer sequences.

UR - http://www.scopus.com/inward/record.url?scp=85063484678&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-12939-2_22

DO - 10.1007/978-3-030-12939-2_22

M3 - Conference contribution

SN - 9783030129385

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 315

EP - 332

BT - Pattern Recognition - 40th German Conference, GCPR 2018, Proceedings

A2 - Fritz, Mario

A2 - Bruhn, Andrés

A2 - Brox, Thomas

PB - Springer-Verlag Italia

ER -