Uma metodologia automática para geração de mapas de disparidades de ambientes reais

• Ano de defesa: 2017
• Autor(a) principal: Carvalho, Carlos Willian de
• Orientador(a): Pedrino, Emerson Carlos
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de São Carlos; Câmpus São Carlos
• Programa de Pós-Graduação: Programa de Pós-Graduação em Ciência da Computação - PPGCC
• Palavras-chave em Português: Visão estéreo; Matriz fundamental; Mapa de disparidades; Autocalibração
• Palavras-chave em Inglês: Stereo vision; Fundamental matrix; Disparity map; Autocalibration
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: https://repositorio.ufscar.br/handle/20.500.14289/9692

The objective of this work was the analysis, combination and adjustments of an algorithm set, in state-of-the-art, building a simple methodology, that is able to represent the depth of the elements in a scene, through the calculation of the disparity map. The map is automatically obtained, from real scenes, with minimal human intervention. Initially, the corners of the objects in a scene are obtained by the SURF algorithm in an image stereo pair, obtained from it. These corners are used to obtain a mapping from points in an image to its correspondences into another image, through a fundamental matrix (obtained by linear estimation, using Least Median of Squares or MSAC); later, the rectification and matching processes are done, generating the map, which suffers a post-processing to fix the distortions. The maps initially obtained was analysed in order to verify the processing time and the similarity with maps obtained by a manual method (which uses calibration). In the future, it is intended to use the studied algorithms as the basis for a system that will allow blind people to do spatial navigation tests in real environments