Evolutionary approaches for endmember extraction in hyperspectral unmixing using genetic algorithms

• Ano de defesa: 2019
• Autor(a) principal: Soares, Douglas Winston Ribeiro
• Orientador(a): Laureano, Gustavo Teodoro
• Banca de defesa: Laureano, Gustavo Teodoro, Coelho, Clarimar Jose, Soares, Anderson da Silva
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• dARK ID: ark:/38995/0013000009c2p
• Idioma: por
• Instituição de defesa: Universidade Federal de Goiás
• Programa de Pós-Graduação: Programa de Pós-graduação em Ciência da Computação (INF)
• Departamento: Instituto de Informática - INF (RG)
• País: Brasil
• Palavras-chave em Português: Desmistura hiperespectral; Extração de endmember; Algoritmo genético; Computação evolucionária
• Palavras-chave em Inglês: Hyperspectral unmixing; Endmember extraction; Genetic algorithm; Evolutionary computing
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: http://repositorio.bc.ufg.br/tede/handle/tede/9999

Endmember Extraction is a critical step in hyperspectral unmixing and classification providing the basis for applications such as identification of minerals, vegetation analysis, geographical survey, disaster management and target identification in military applications. The endemember extraction determines the basic constituent materials contained in the hyperspectral image while providing the requirements to the abundance inversion process, used to obtain the percentage of several endmembers in each pixel. Nevertheless, low spatial resolution and computing time are the two major difficulties, the first due to the spatial interactions of different fractions of mixed endmembers and the second due to strict and extensive search utilized in state-of-the-art methods. Three evolutionary endmember extractors are proposed, so-called GAEE, GAEEIVFm and GAEEII. The first is based on solving a linear endmember extraction problem as an evolutionary optimization task, maximizing the simplex volume in the endmember search space, GAEE-IVFm represents a variation with of the GAEE with an In Vitro Fertilization module, and the GAEEII is a multi-epochs genetic algorithm with enhancements to the naive genetic algorithm endmember extractor (GAEE). To demonstrate the superiority of the proposed methods, extensive experiments are conducted on several well-known real and synthetic hyperspectral images, as well as a possible relationship between the spectral angle distance (SAD) and the volume of the simplex. The results confirm that the proposed methods considerably improved, up to three times increase in accuracy and scalable computing time compared to the state-of-the-art techniques in the literature including recent developments.