Sistema de reconhecimento biométrico baseado em marcha utilizando poses 2D
Carregando...
Data
2022-03-03
Autores
Jangua, Daniel Ricardo dos Santos
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
Nas últimas décadas a biometria, utilizada para identificação de indivíduos por meio do estudo de suas características físicas e comportamentais, tem se tornado uma ferramenta cada vez mais presente no cotidiano da sociedade, sendo aplicada em cenários cada vez mais desafiadores. Apesar disso, ainda existe um importante desafio a ser superado: a identificação automática de indivíduos em vídeos de baixa resolução, à distância, de maneira sigilosa e não invasiva. Nesse cenário, em que os sistemas biométricos convencionais não se mostram eficazes, o reconhecimento de marcha ganha destaque. Entretanto, mesmo os métodos de reconhecimento de marcha considerados estado-da-arte ainda não apresentam a robustez adequada para serem utilizados em situações reais, para solucionar essa questão, pode-se adotar técnicas de multibiometria, que buscam combinar diferentes métodos biométricos em um único sistema. Neste trabalho, foram estudados e avaliados dois métodos de reconhecimento de marcha baseados em poses 2D e um método baseado em medidas antropométricas, bem como a fusão entre esses métodos utilizando multibiometria. O resultado dos experimentos mostra que, no geral, a fusão entre os métodos apresentam melhorias consideráveis em termos de acurácia. Por fim, foi desenvolvido um sistema biométrico baseado no reconhecimento de marcha utilizando a fusão de métodos que apresentou o melhor resultado.
In the last decades, biometrics, used to identify individuals through the study of their physical and behavioral characteristics, has become an increasingly present tool in society's daily life, being applied in increasingly challenging scenarios. Despite this, there is still an important challenge to be overcome: the automatic identification of individuals in low resolution videos, at a distance, in a confidential and non-invasive way. In this scenario, in which conventional biometric systems are not effective, gait recognition is highlighted. However, even state-of-the-art gait recognition methods still do not have the adequate robustness to be used in real situations, to solve this issue, multibiometric techniques can be adopted, which seek to combine different biometric methods into a single system. In this work, two gait recognition methods based on 2D poses and a method based on anthropometric measurements were studied and evaluated, as well as the fusion between these methods using multibiometrics. The results of the experiments show that, in general, the fusion between the methods presents considerable improvements in terms of accuracy. Finally, a biometric system based on gait recognition was developed using the fusion of methods that presented the best result.
In the last decades, biometrics, used to identify individuals through the study of their physical and behavioral characteristics, has become an increasingly present tool in society's daily life, being applied in increasingly challenging scenarios. Despite this, there is still an important challenge to be overcome: the automatic identification of individuals in low resolution videos, at a distance, in a confidential and non-invasive way. In this scenario, in which conventional biometric systems are not effective, gait recognition is highlighted. However, even state-of-the-art gait recognition methods still do not have the adequate robustness to be used in real situations, to solve this issue, multibiometric techniques can be adopted, which seek to combine different biometric methods into a single system. In this work, two gait recognition methods based on 2D poses and a method based on anthropometric measurements were studied and evaluated, as well as the fusion between these methods using multibiometrics. The results of the experiments show that, in general, the fusion between the methods presents considerable improvements in terms of accuracy. Finally, a biometric system based on gait recognition was developed using the fusion of methods that presented the best result.
Descrição
Palavras-chave
Estimação de Pose, Biometria, Marcha humana