Monitoramento de vegetação invasora através das épocas com UAV e Deep Learning

• Ano de defesa: 2022
• Autor(a) principal: Camargo, Charles Pires de
• Orientador(a): Vivaldini, Kelen Cristiane Teixeira
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de São Carlos; Câmpus Sorocaba
• Programa de Pós-Graduação: Programa de Pós-Graduação em Ciência da Computação - PPGCC-So
• Palavras-chave em Português: Aprendizagem de máquina; UAV; Segmentação semântica
• Palavras-chave em Inglês: Deep learning; Semantic segmentation
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO::SISTEMAS DE COMPUTACAO
• Link de acesso: https://repositorio.ufscar.br/handle/ufscar/15883

Species originating from one biome are often irregularly introduced in other biomes, even if unintentionally, with the sole desire of their own consumption or for ornamental purposes. However, these species can end up becoming invasive and subjugating native vegetation. In this scenario, we have characterized biological invasion, which can cause irreversible negative impacts on biodiversity and affect economic productivity in sectors such as fisheries, forestry and agriculture. Furthermore, many species are vectors of human diseases, making biological invasions a major problem. With many closed forests, regions such as the mountains of the sea, and other places that are difficult to access, monitoring the Brazilian territory becomes very difficult and requires many resources for your care, whether human or financial. Remotely and automatically detecting invasive vegetation in large regions or areas of difficult physical access can be a very positive factor for conservation work. Through this monitoring, concrete actions can be taken in favor of the environment and irreversible damage to the ecosystem can be avoided. Making use of Deep Learning models for the detection of the invasive species Hedychium coronarium in images obtained through remote sensing, this master's project proposes a methodology for the monitoring over time of the area invaded by Hedychium coronarium in order to help specialists in answers to questions ecological