• Login
    View Item 
    •   DSpace Home
    • Trabalhos de Graduação
    • Engenharia Agrícola (Jandaia do Sul)
    • View Item
    •   DSpace Home
    • Trabalhos de Graduação
    • Engenharia Agrícola (Jandaia do Sul)
    • View Item
    JavaScript is disabled for your browser. Some features of this site may not work without it.

    Mapeamento do uso e ocupação do solo por meio da plataforma Google Earth Engine utilizando dados Sentinel-2 e SRTM

    Thumbnail
    View/Open
    M_DANIEL_FELIPE_LUSCENTI.pdf (3.399Mb)
    Date
    2022
    Author
    Luscenti, Daniel Felipe
    Metadata
    Show full item record
    Abstract
    Resumo: O sensoriamento remoto desempenha papel relevante na análise de imagens de satélite e, quando associado a técnicas e ferramentas de machine learning, possibilita a extração de informações úteis à tomada de decisão, com potencial impacto positivo sobre a agricultura brasileira. Nesse contexto, o objetivo do presente trabalho foi determinar o uso e a ocupação do solo na região do Vale do Ivaí (PR) por meio do classificador Random Forest, aplicado na plataforma Google Earth Engine (GEE) com imagens do satélite Sentinel-2. Adicionalmente, avaliou-se o efeito da incorporação da variável declividade, obtida a partir de dados do Shuttle Radar Topography Mission (SRTM), no desempenho da classificação. Para tanto, desenvolveu-se um algoritmo em linguagem JavaScript no ambiente do GEE para classificar as imagens Sentinel-2 na área de estudo, discriminando as classes: vegetação florestal, solo exposto, corpos hídricos, pastagem, agricultura e edificações. Posteriormente, verificou-se semelhança entre a distribuição espacial dos usos e ocupações do solo e o padrão de declividade, especialmente para as classes agricultura e pastagem, que apresentaram maior predominância e acompanharam as tendências observadas no mapa de declividade. A acurácia global das classificações, com e sem a inclusão da declividade, foi considerada excelente, de acordo com os critérios usualmente adotados para o índice Kappa (0,92 e 0,91, respectivamente) e para a exatidão global (94,94% e 94,29%, respectivamente). Esses resultados indicam que a inclusão da declividade contribuiu positivamente para o processo de classificação, ainda que com ganhos marginais
     
    Abstract: Remote sensing plays a key role in the analysis of satellite imagery and, when combined with machine learning techniques, enables the extraction of information that supports decision-making with potential benefits for agriculture. This study aimed to map land use and land cover (LULC) in the Vale do Ivaí region, Paraná State, Brazil, using a Random Forest classifier implemented in the Google Earth Engine (GEE) platform with Sentinel-2 imagery. Additionally, the influence of incorporating slope information derived from Shuttle Radar Topography Mission (SRTM) data on classification performance was assessed. A JavaScript-based algorithm was developed in the GEE environment to classify Sentinel-2 images into six classes: forest vegetation, bare soil, water bodies, pasture, agriculture, and built-up areas. The spatial distribution of the mapped classes showed similarity to the regional slope pattern, particularly for agriculture and pasture, the dominant classes, which followed the main trends observed in the slope map. Classification accuracy was considered excellent both with and without slope information, according to the Kappa index (0.92 and 0.91, respectively) and overall accuracy (94.94% and 94.29%, respectively). These results indicate that slope contributed positively to the classification process, albeit with marginal gains
     
    URI
    https://hdl.handle.net/1884/105785
    Collections
    • Engenharia Agrícola (Jandaia do Sul) [33]

    DSpace software copyright © 2002-2022  LYRASIS
    Contact Us | Send Feedback
    Theme by 
    Atmire NV
     

     

    Browse

    All of DSpaceCommunities & CollectionsBy Issue DateAuthorsTitlesSubjectsxmlui.ArtifactBrowser.Navigation.browse_typeThis CollectionBy Issue DateAuthorsTitlesSubjectsxmlui.ArtifactBrowser.Navigation.browse_type

    My Account

    LoginRegister

    Statistics

    View Usage Statistics

    DSpace software copyright © 2002-2022  LYRASIS
    Contact Us | Send Feedback
    Theme by 
    Atmire NV