3D hypothesis clustering for cross-view matching in multi-person motion capture
Abstract We present a multiview method for markerless motion capture of multiple people. The main challenge in this problem is to determine cross-view correspondences for the 2D joints in the presence of noise. We propose a 3D hypothesis clustering technique to solve this problem. The core idea is t...
Main Authors: | , , |
---|---|
Format: | Article |
Language: | English |
Published: |
SpringerOpen
2020-06-01
|
Series: | Computational Visual Media |
Subjects: | |
Online Access: | http://link.springer.com/article/10.1007/s41095-020-0171-y |
id |
doaj-126b0ea5f44a48198a91b0dc390aaa56 |
---|---|
record_format |
Article |
spelling |
doaj-126b0ea5f44a48198a91b0dc390aaa562020-11-25T03:37:41ZengSpringerOpenComputational Visual Media2096-04332096-06622020-06-016214715610.1007/s41095-020-0171-y3D hypothesis clustering for cross-view matching in multi-person motion captureMiaopeng Li0Zimeng Zhou1Xinguo Liu2State Key Lab of CAD&CG, Zhejiang UniversityState Key Lab of CAD&CG, Zhejiang UniversityState Key Lab of CAD&CG, Zhejiang UniversityAbstract We present a multiview method for markerless motion capture of multiple people. The main challenge in this problem is to determine cross-view correspondences for the 2D joints in the presence of noise. We propose a 3D hypothesis clustering technique to solve this problem. The core idea is to transform joint matching in 2D space into a clustering problem in a 3D hypothesis space. In this way, evidence from photometric appearance, multiview geometry, and bone length can be integrated to solve the clustering problem efficiently and robustly. Each cluster encodes a set of matched 2D joints for the same person across different views, from which the 3D joints can be effectively inferred. We then assemble the inferred 3D joints to form full-body skeletons for all persons in a bottom–up way. Our experiments demonstrate the robustness of our approach even in challenging cases with heavy occlusion, closely interacting people, and few cameras. We have evaluated our method on many datasets, and our results show that it has significantly lower estimation errors than many state-of-the-art methods.http://link.springer.com/article/10.1007/s41095-020-0171-ymulti-person motion capturecross-view matchingclusteringhuman pose estimation |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Miaopeng Li Zimeng Zhou Xinguo Liu |
spellingShingle |
Miaopeng Li Zimeng Zhou Xinguo Liu 3D hypothesis clustering for cross-view matching in multi-person motion capture Computational Visual Media multi-person motion capture cross-view matching clustering human pose estimation |
author_facet |
Miaopeng Li Zimeng Zhou Xinguo Liu |
author_sort |
Miaopeng Li |
title |
3D hypothesis clustering for cross-view matching in multi-person motion capture |
title_short |
3D hypothesis clustering for cross-view matching in multi-person motion capture |
title_full |
3D hypothesis clustering for cross-view matching in multi-person motion capture |
title_fullStr |
3D hypothesis clustering for cross-view matching in multi-person motion capture |
title_full_unstemmed |
3D hypothesis clustering for cross-view matching in multi-person motion capture |
title_sort |
3d hypothesis clustering for cross-view matching in multi-person motion capture |
publisher |
SpringerOpen |
series |
Computational Visual Media |
issn |
2096-0433 2096-0662 |
publishDate |
2020-06-01 |
description |
Abstract We present a multiview method for markerless motion capture of multiple people. The main challenge in this problem is to determine cross-view correspondences for the 2D joints in the presence of noise. We propose a 3D hypothesis clustering technique to solve this problem. The core idea is to transform joint matching in 2D space into a clustering problem in a 3D hypothesis space. In this way, evidence from photometric appearance, multiview geometry, and bone length can be integrated to solve the clustering problem efficiently and robustly. Each cluster encodes a set of matched 2D joints for the same person across different views, from which the 3D joints can be effectively inferred. We then assemble the inferred 3D joints to form full-body skeletons for all persons in a bottom–up way. Our experiments demonstrate the robustness of our approach even in challenging cases with heavy occlusion, closely interacting people, and few cameras. We have evaluated our method on many datasets, and our results show that it has significantly lower estimation errors than many state-of-the-art methods. |
topic |
multi-person motion capture cross-view matching clustering human pose estimation |
url |
http://link.springer.com/article/10.1007/s41095-020-0171-y |
work_keys_str_mv |
AT miaopengli 3dhypothesisclusteringforcrossviewmatchinginmultipersonmotioncapture AT zimengzhou 3dhypothesisclusteringforcrossviewmatchinginmultipersonmotioncapture AT xinguoliu 3dhypothesisclusteringforcrossviewmatchinginmultipersonmotioncapture |
_version_ |
1724544484874649600 |