Clustering satellite image time series data based on growing self-organizing maps

• Ano de defesa: 2020
• Autor(a) principal: Rodrigo de Sales da Silva Adeu
• Orientador(a): Karine Reis Ferreira, Pedro Ribeiro de Andrade Neto
• Banca de defesa: Michelle Cristina Araujo Picoli, Luciana Alvim Santos Romani
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Instituto Nacional de Pesquisas Espaciais (INPE)
• Programa de Pós-Graduação: Programa de Pós-Graduação do INPE em Computação Aplicada
• País: BR
• Link de acesso: http://urlib.net/sid.inpe.br/mtc-m21c/2020/08.03.21.55

Mapping Earth land use and land cover is crucial to understand agricultural dynamics. Recently, analysis of time series extracted from Earth observation satellite images has been widely used to produce land use and land cover information. In time series analysis, clustering is a common technique performed to discover patterns on data sets. In this work, we evaluate the Growing Self-Organizing Maps (GSOMs) algorithm for clustering satellite image time series and compare it with Self-Organizing Maps (SOMs) algorithm. This paper presents two case studies using satellite image time series associated to samples of land use and land cover classes, highlighting the advantage of providing a neutral factor (called spread factor) as a parameter for GSOM, instead of the SOM grid size. We first compare GSOM with traditional SOM, analyzing the resultant network topology, the algorithm running time, the cluster accuracy and the neighborhood maintenance. In the second case study, we changed the dataset, increasing the number of samples and repeating the analysis. We finish concluding that it is possible to cluster satellite image time series with GSOM, avoiding the SOM grid size additional parameter. Besides that, GSOM keeps most of SOM properties and can be considered as a suitable alternative to SOM.