Baptizo: a sensor fusion based model for tracking the identity of human poses
Autor(a) principal: | |
---|---|
Data de Publicação: | 2019 |
Tipo de documento: | Dissertação |
Idioma: | eng |
Título da fonte: | Repositório Institucional da UNISINOS (RBDU Repositório Digital da Biblioteca da Unisinos) |
Texto Completo: | http://www.repositorio.jesuita.org.br/handle/UNISINOS/9078 |
Resumo: | Recent advances in the capabilities of computing devices enable new methods to estimate the pose of humans. Human pose estimation techniques are relevant for several industry fields, such as surveillance and interactive entertainment. Further, encoded human poses provide a valuable input for behavioral analysis and activity recognition. Body part detectors offer millimetric accuracy thanks to state-of-the-art Computer Vision technology. However, they still suffer from issues, such as long-term occlusion, that hinder the identification of human subjects. Such problems are intrinsic to Computer Vision devices and can only be solved either with the use of heuristic methods or the deployment of more cameras, which are not always feasible. In turn, radiofrequency-based tracking systems do not suffer from occlusion or identity loss problems and, albeit not as precise as Computer Vision methods, can achieve a high accuracy level. Radiofrequency positioning systems and human pose estimation techniques can complement each other in different ways. For example, the prior can help to identify tracked humans and reduce occlusion errors while the later can increase the accuracy of obtained positions. Thus, the combination of radiofrequency-based positioning and computer vision-based human pose estimation yields a solution that provides better tracking results. Therefore, this thesis proposes a system that generates identified pose data by fusing the unique identities of radiofrequency sensors with unidentified body poses while using estimated body parts for reducing radiofrequency position estimations errors. Experiments with a proof-of-concept demonstrate the feasibility of the sensor fusion technique. Furthermore, experiments analyzing the proposed error reductiong strategy conducted in a experimentation laboratory and a real operating room also show a potential reduction on positioning errors by nearly 46%. |
id |
USIN_5f6a0fb9aa516398e966a15adcf58c40 |
---|---|
oai_identifier_str |
oai:www.repositorio.jesuita.org.br:UNISINOS/9078 |
network_acronym_str |
USIN |
network_name_str |
Repositório Institucional da UNISINOS (RBDU Repositório Digital da Biblioteca da Unisinos) |
repository_id_str |
|
spelling |
2020-02-21T14:20:39Z2020-02-21T14:20:39Z2019-08-13Submitted by JOSIANE SANTOS DE OLIVEIRA (josianeso) on 2020-02-21T14:20:38Z No. of bitstreams: 1 Rodrigo Bazo_.pdf: 6143424 bytes, checksum: 3e12a11bca2077ba38dd28af72018915 (MD5)Made available in DSpace on 2020-02-21T14:20:39Z (GMT). No. of bitstreams: 1 Rodrigo Bazo_.pdf: 6143424 bytes, checksum: 3e12a11bca2077ba38dd28af72018915 (MD5) Previous issue date: 2019-08-13Recent advances in the capabilities of computing devices enable new methods to estimate the pose of humans. Human pose estimation techniques are relevant for several industry fields, such as surveillance and interactive entertainment. Further, encoded human poses provide a valuable input for behavioral analysis and activity recognition. Body part detectors offer millimetric accuracy thanks to state-of-the-art Computer Vision technology. However, they still suffer from issues, such as long-term occlusion, that hinder the identification of human subjects. Such problems are intrinsic to Computer Vision devices and can only be solved either with the use of heuristic methods or the deployment of more cameras, which are not always feasible. In turn, radiofrequency-based tracking systems do not suffer from occlusion or identity loss problems and, albeit not as precise as Computer Vision methods, can achieve a high accuracy level. Radiofrequency positioning systems and human pose estimation techniques can complement each other in different ways. For example, the prior can help to identify tracked humans and reduce occlusion errors while the later can increase the accuracy of obtained positions. Thus, the combination of radiofrequency-based positioning and computer vision-based human pose estimation yields a solution that provides better tracking results. Therefore, this thesis proposes a system that generates identified pose data by fusing the unique identities of radiofrequency sensors with unidentified body poses while using estimated body parts for reducing radiofrequency position estimations errors. Experiments with a proof-of-concept demonstrate the feasibility of the sensor fusion technique. Furthermore, experiments analyzing the proposed error reductiong strategy conducted in a experimentation laboratory and a real operating room also show a potential reduction on positioning errors by nearly 46%.Os recentes avanços no poder computacional de dispositivos permitem a utilização de novos métodos para a estimativa de poses humanas. Tais técnicas são relevantes para diversos setores da indústria, como segurança e entretenimento. Além disso, poses humanas são um input valioso para análise comportamental e reconhecimento de atividades. Reconhecedores de partes de corpo humana, utilizados em estimativas de pose humana, possuem precisão milimétrica devido aos equipamentos de estado da arte de visão computacional. Porém, estes equipamentos possuem limitações como a oclusão, que dificulta a identificação de pessoas. Tais problemas são nativos aos dispositivos de visão computacional devido a sua natureza, e somente podem ser superados utilizando heuristicas ou aumentando o numero de câmeras, o que não é sempre viável. Por outro lado, sistemas de rastreamento baseados em radiofrequência não sofrem com oclusão ou problemas como perda de identidade, e também alcançam altos níveis de precisão mesmo não sendo tão precisos quanto métodos de visão computacional. Sistemas de rastreamento baseados em radiofrequência e estimativas de pose humanas podem se complementar de diversas maneirars. Por exemplo, o primeiro pode ajudar na identificação de poses estimadas, e as poses podem ser utilizadas para mitigar os erros obtidos. Desta maneira, a combinação de ambas as tecnologias oferecem um resultado de rastreamento de poses com precisão superior. Esta dissertação propõem um sistema que gera poses identificadas, baseado na fusão de identificadores de radiofrequência com poses obtidas através de técnicas de visão computacional. Além disso, uma técnica para redução de erro na estimativa da posição dos dispositivos de radiofrequência utilizando poses estimadas é proposta. Experimentos demonstram a viabilidade da fusão de ambos tipos de dados. Além disso, reduções de erros de até 46% utilizando a estratégia de redução de erro proposta são observados. Tanto em experimentos conduzidos em um laboratório de experimentação quanto em uma sala cirúrgica real.UNISINOS - Universidade do Vale do Rio dos SinosBazo, Rodrigohttp://lattes.cnpq.br/9476895751212336http://lattes.cnpq.br/9637121030877187Costa, Cristiano André daUniversidade do Vale do Rio dos SinosPrograma de Pós-Graduação em Computação AplicadaUnisinosBrasilEscola PolitécnicaBaptizo: a sensor fusion based model for tracking the identity of human posesACCNPQ::Ciências Exatas e da Terra::Ciência da ComputaçãoRastreamentoRadiofrequênciaVisão ComputacionalEstimativa de Pose HumanaSensor FusionTrackingRadio FrequencyComputer VisionHuman Pose EstimationSensor Fusioninfo:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/masterThesishttp://www.repositorio.jesuita.org.br/handle/UNISINOS/9078info:eu-repo/semantics/openAccessengreponame:Repositório Institucional da UNISINOS (RBDU Repositório Digital da Biblioteca da Unisinos)instname:Universidade do Vale do Rio dos Sinos (UNISINOS)instacron:UNISINOSORIGINALRodrigo Bazo_.pdfRodrigo Bazo_.pdfapplication/pdf6143424http://repositorio.jesuita.org.br/bitstream/UNISINOS/9078/1/Rodrigo+Bazo_.pdf3e12a11bca2077ba38dd28af72018915MD51LICENSElicense.txtlicense.txttext/plain; charset=utf-82175http://repositorio.jesuita.org.br/bitstream/UNISINOS/9078/2/license.txt320e21f23402402ac4988605e1edd177MD52UNISINOS/90782020-02-21 11:23:03.023oai:www.repositorio.jesuita.org.br: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 Digital de Teses e Dissertaçõeshttp://www.repositorio.jesuita.org.br/oai/requestopendoar:2020-02-21T14:23:03Repositório Institucional da UNISINOS (RBDU Repositório Digital da Biblioteca da Unisinos) - Universidade do Vale do Rio dos Sinos (UNISINOS)false |
dc.title.pt_BR.fl_str_mv |
Baptizo: a sensor fusion based model for tracking the identity of human poses |
title |
Baptizo: a sensor fusion based model for tracking the identity of human poses |
spellingShingle |
Baptizo: a sensor fusion based model for tracking the identity of human poses Bazo, Rodrigo ACCNPQ::Ciências Exatas e da Terra::Ciência da Computação Rastreamento Radiofrequência Visão Computacional Estimativa de Pose Humana Sensor Fusion Tracking Radio Frequency Computer Vision Human Pose Estimation Sensor Fusion |
title_short |
Baptizo: a sensor fusion based model for tracking the identity of human poses |
title_full |
Baptizo: a sensor fusion based model for tracking the identity of human poses |
title_fullStr |
Baptizo: a sensor fusion based model for tracking the identity of human poses |
title_full_unstemmed |
Baptizo: a sensor fusion based model for tracking the identity of human poses |
title_sort |
Baptizo: a sensor fusion based model for tracking the identity of human poses |
author |
Bazo, Rodrigo |
author_facet |
Bazo, Rodrigo |
author_role |
author |
dc.contributor.authorLattes.pt_BR.fl_str_mv |
http://lattes.cnpq.br/9476895751212336 |
dc.contributor.advisorLattes.pt_BR.fl_str_mv |
http://lattes.cnpq.br/9637121030877187 |
dc.contributor.author.fl_str_mv |
Bazo, Rodrigo |
dc.contributor.advisor1.fl_str_mv |
Costa, Cristiano André da |
contributor_str_mv |
Costa, Cristiano André da |
dc.subject.cnpq.fl_str_mv |
ACCNPQ::Ciências Exatas e da Terra::Ciência da Computação |
topic |
ACCNPQ::Ciências Exatas e da Terra::Ciência da Computação Rastreamento Radiofrequência Visão Computacional Estimativa de Pose Humana Sensor Fusion Tracking Radio Frequency Computer Vision Human Pose Estimation Sensor Fusion |
dc.subject.por.fl_str_mv |
Rastreamento Radiofrequência Visão Computacional Estimativa de Pose Humana |
dc.subject.spa.fl_str_mv |
Sensor Fusion Tracking Radio Frequency Computer Vision Human Pose Estimation Sensor Fusion |
description |
Recent advances in the capabilities of computing devices enable new methods to estimate the pose of humans. Human pose estimation techniques are relevant for several industry fields, such as surveillance and interactive entertainment. Further, encoded human poses provide a valuable input for behavioral analysis and activity recognition. Body part detectors offer millimetric accuracy thanks to state-of-the-art Computer Vision technology. However, they still suffer from issues, such as long-term occlusion, that hinder the identification of human subjects. Such problems are intrinsic to Computer Vision devices and can only be solved either with the use of heuristic methods or the deployment of more cameras, which are not always feasible. In turn, radiofrequency-based tracking systems do not suffer from occlusion or identity loss problems and, albeit not as precise as Computer Vision methods, can achieve a high accuracy level. Radiofrequency positioning systems and human pose estimation techniques can complement each other in different ways. For example, the prior can help to identify tracked humans and reduce occlusion errors while the later can increase the accuracy of obtained positions. Thus, the combination of radiofrequency-based positioning and computer vision-based human pose estimation yields a solution that provides better tracking results. Therefore, this thesis proposes a system that generates identified pose data by fusing the unique identities of radiofrequency sensors with unidentified body poses while using estimated body parts for reducing radiofrequency position estimations errors. Experiments with a proof-of-concept demonstrate the feasibility of the sensor fusion technique. Furthermore, experiments analyzing the proposed error reductiong strategy conducted in a experimentation laboratory and a real operating room also show a potential reduction on positioning errors by nearly 46%. |
publishDate |
2019 |
dc.date.issued.fl_str_mv |
2019-08-13 |
dc.date.accessioned.fl_str_mv |
2020-02-21T14:20:39Z |
dc.date.available.fl_str_mv |
2020-02-21T14:20:39Z |
dc.type.status.fl_str_mv |
info:eu-repo/semantics/publishedVersion |
dc.type.driver.fl_str_mv |
info:eu-repo/semantics/masterThesis |
format |
masterThesis |
status_str |
publishedVersion |
dc.identifier.uri.fl_str_mv |
http://www.repositorio.jesuita.org.br/handle/UNISINOS/9078 |
url |
http://www.repositorio.jesuita.org.br/handle/UNISINOS/9078 |
dc.language.iso.fl_str_mv |
eng |
language |
eng |
dc.rights.driver.fl_str_mv |
info:eu-repo/semantics/openAccess |
eu_rights_str_mv |
openAccess |
dc.publisher.none.fl_str_mv |
Universidade do Vale do Rio dos Sinos |
dc.publisher.program.fl_str_mv |
Programa de Pós-Graduação em Computação Aplicada |
dc.publisher.initials.fl_str_mv |
Unisinos |
dc.publisher.country.fl_str_mv |
Brasil |
dc.publisher.department.fl_str_mv |
Escola Politécnica |
publisher.none.fl_str_mv |
Universidade do Vale do Rio dos Sinos |
dc.source.none.fl_str_mv |
reponame:Repositório Institucional da UNISINOS (RBDU Repositório Digital da Biblioteca da Unisinos) instname:Universidade do Vale do Rio dos Sinos (UNISINOS) instacron:UNISINOS |
instname_str |
Universidade do Vale do Rio dos Sinos (UNISINOS) |
instacron_str |
UNISINOS |
institution |
UNISINOS |
reponame_str |
Repositório Institucional da UNISINOS (RBDU Repositório Digital da Biblioteca da Unisinos) |
collection |
Repositório Institucional da UNISINOS (RBDU Repositório Digital da Biblioteca da Unisinos) |
bitstream.url.fl_str_mv |
http://repositorio.jesuita.org.br/bitstream/UNISINOS/9078/1/Rodrigo+Bazo_.pdf http://repositorio.jesuita.org.br/bitstream/UNISINOS/9078/2/license.txt |
bitstream.checksum.fl_str_mv |
3e12a11bca2077ba38dd28af72018915 320e21f23402402ac4988605e1edd177 |
bitstream.checksumAlgorithm.fl_str_mv |
MD5 MD5 |
repository.name.fl_str_mv |
Repositório Institucional da UNISINOS (RBDU Repositório Digital da Biblioteca da Unisinos) - Universidade do Vale do Rio dos Sinos (UNISINOS) |
repository.mail.fl_str_mv |
|
_version_ |
1801845038610644992 |