Reconstrução de um modelo 3D a partir de imagens térmicas 2D de uma mama via câmera infravermelha

• Ano de defesa: 2021
• Autor(a) principal: Costa, Guilherme Ramalho
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Uberlândia; Brasil; Programa de Pós-graduação em Engenharia Mecânica
• Palavras-chave em Português: Visão estéreo; Geometria epipolar; Termogramas; Câncer de mama; Stereo vision; Epipolar geometry; Thermograms; Breast cancer; CNPQ::ENGENHARIAS::ENGENHARIA MECANICA; Visão estereoscópica; Mamas - Câncer
• Link de acesso: https://repositorio.ufu.br/handle/123456789/31279; http://doi.org/10.14393/ufu.di.2021.6006

With the development of image processing algorithms and the arrival of thermal cameras, a new concept of temperature analysis emerged. Besides industrial processes, thermal images also started to play an important role in the evaluation of the temperature of human beings, being used for the diagnosis of diseases and other anomalies in which the body undergoes a change in its considered normal temperature. Having a metabolism different from the cells of the human body, the tumor cells cause the temperature of the region where they are to be, also, different from that found in non-tumor cells. This work aims to develop three-dimensional models with temperature information from two-dimensional thermal. With the three-dimensional thermal maps, we can have a closer view of the real geometries found in the images and, with this, new information for the application of the bio-heat transfer equations in numeric simulations also involving geometric parameters of the breast. Using the stereoscopic view, which is based on epipolar geometry, it was possible to develop a method of generating representative dense points clouds from thermal images, as well as verifying the developed method particularities and limitations. The developed three-dimensional maps showed well-defined and visible characteristics, although the input images had low spatial resolution, which limited the application of the method.