Powered by OpenAIRE graph
Found an issue? Give us feedback
addClaim

Reconhecimento facial em ambiente não cooperativo

Authors: Menino, Rúben Miguel Paulo;

Reconhecimento facial em ambiente não cooperativo

Abstract

Nos dias de hoje, o reconhecimento facial tornou-se uma marco bastante importante na área da informática e tem vindo a receber bastante atenção ao longo dos anos. O reconhecimento facial pode ser utilizado em bastantes áreas, porém uma área que tem estado em bastante crescimento é a área da segurança. Temas como acessos a instalações militares, identificação de grupos terroristas e pessoas que forçam e abusam da lei são alguns dos temas mais abordados. Apesar de ser um tema bastante estudado existem ainda algumas limitações, principalmente quando a aquisição das imagem são adquiridas de pessoas em ambientes não-cooperativos. O objetivo desta dissertação de mestrado é a investigação de vários métodos de deteção e reconhecimento facial. Apresenta um estudo sobre os mais importantes algoritmos, e técnicas de pré-processamento como frontalização e alinhamento facial de modo a conseguir comparar os níveis de precisão de cada um dos algoritmos. De modo a obter resultados foi efetuado um dataset de imagens na Universidade de Aveiro de vários espetros de cores. Foi possível observar que algoritmos baseados em redes neurais convolucional profundas têm uma precisão mais elevada em relação a vários métodos tradicionais. Foi ainda dado um primeiro passo no sentido de desenvolver um modelo de deteção facial em imagens térmicas, onde existiu uma melhoria de cerca de 30% em relação ao modelo original.

Nowadays, facial recognition has become very important in the field of computing and has been receiving a lot of attention over the years. Facial recognition can be used in many areas, but one area that has been growing a lot is security. Topics like access to military installations, identification of terrorist groups and people who force and abuse the law are some of the most discussed topics. Despite being a widely studied topic, there are still some limitations, especially when image acquisition is acquired from people in non-cooperative environments. The objective of this master’s thesis is the investigation of various methods of detection and facial recognition. It presents a study on the most important algorithms, and preprocessing techniques such as frontalization and facial alignment in order to be able to compare the accuracy levels of each of the algorithms. In order to obtain results, an image dataset was carried out at the University of Aveiro of various color spectrum. It was possible to observe that algorithms based on deep convolutional neural networks have a higher precision compared to several traditional methods. A first step was also taken towards developing a model of facial detection in thermal images, where there was an improvement of about 30% compared to the original model.

Mestrado em Engenharia de Computadores e Telemática

Country
Portugal
Related Organizations
Keywords

Image processing, Image acquisition, Deep learning, Face detection, Face recognition, Algorithms

  • BIP!
    Impact byBIP!
    selected citations
    These citations are derived from selected sources.
    This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    0
    popularity
    This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
    Average
    influence
    This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    Average
    impulse
    This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
    Average
Powered by OpenAIRE graph
Found an issue? Give us feedback
selected citations
These citations are derived from selected sources.
This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Citations provided by BIP!
popularity
This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
BIP!Popularity provided by BIP!
influence
This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Influence provided by BIP!
impulse
This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
BIP!Impulse provided by BIP!
0
Average
Average
Average
Upload OA version
Are you the author of this publication? Upload your Open Access version to Zenodo!
It’s fast and easy, just two clicks!