• Entrar
    Ver item 
    •   Página inicial
    • BIBLIOTECA DIGITAL: Trabalhos de Graduação
    • Engenharia Agrícola (Jandaia do Sul)
    • Ver item
    •   Página inicial
    • BIBLIOTECA DIGITAL: Trabalhos de Graduação
    • Engenharia Agrícola (Jandaia do Sul)
    • Ver item
    JavaScript is disabled for your browser. Some features of this site may not work without it.

    Padrões de transição da terra no Vale do lvaí-PR identificadas por técnicas de machine learning

    Thumbnail
    Visualizar/Abrir
    M_ROBERTO_DE_SOUZA_JUNIOR.pdf (5.765Mb)
    Data
    2021
    Autor
    Souza Junior, Roberto de Assis de
    Metadata
    Mostrar registro completo
    Resumo
    Resumo: O Brasil é considerado um dos maiores produtores e exportadores de alimentos do mundo, sendo o maior produtor e o segundo maior exportador de soja, e o maior produtor e exportador de derivados da cana-de-açúcar. Devido ao aumento da demanda mundial de alimentos, a competição por área entre algumas commodities motivadas principalmente pela valorização dos preços, vem resultando em mudanças no uso e cobertura do solo. Pesquisas e estudos direcionados a respeito da cobertura e quantificação do uso do solo tornam-se fundamentais para diversos ramos do conhecimento que analisam as variações ocorridas na área. Neste propósito, a utilização de técnicas de sensoriamento remoto apresenta grandes vantagens, como a possibilidade de recobrimento de grandes extensões de áreas em tempo hábil. Deste modo, esse trabalho teve como objetivo determinar o uso e ocupação do solo em 8 categorias (Mata Nativa, Reflorestamento, Corpos Hídricos, Solo Exposto, Pastagem, Milho, Cana-de-açúcar e Soja) no território do Vale do Ivaí – PR por meio de sensoriamento remoto e algoritmos de machine learning, e identificar os padrões de transições do solo ocorridos no período analisado através da utilização da metodologia intensity analysis. Para realizar as classificações do território estudado, foram treinados e testados dois modelos de machine learning, uma rede neural artificial rasa (SNN) e uma rede neural artificial profunda, popularmente conhecida como deep learning (DNN). De acordo com os resultados obtidos podemos dizer que, o algoritmo SNN foi o que melhor se destacou, apresentando os valores de exatidão global e Índice Kappa, 98,95% e 98,77%, respectivamente. Através da aplicação da metodologia intensity analysis, identificou-se que a categoria pastagem foi a que mais perdeu área, e que a cana-de-açúcar foi a categoria que mais tomou as áreas, em outras palavras, podemos dizer que cana-de-açúcar foi a cultura que mais avançou de maneira intensa sobre as áreas de pastagem. Podemos concluir que esse padrão de transição ocorreu devido a agricultura e a cultura da cana-de-açúcar em específico, tender a se expandir sobre as áreas com baixas produtividades, e/ou com baixos rendimentos
     
    Abstract: Brazil is considered one of the largest producers and exporters of food in the world, being the largest producer and the second largest exporter of soybeans, and the largest producer and exporter of sugar cane derivatives. Due to the increase in world demand for food, competition by area among some commodities, mainly motivated by price appreciation, has resulted in changes in land use and cover. Researches and studies directed towards the coverage and quantification of the use of the land become fundamental for diverse branches of the knowledge that analyze the variations occurred in the area. In this regard, the use of remote sensing techniques has great advantages, such as the possibility of covering large areas in a timely manner. Thus, this work aimed to determine the use and occupation of the soil in 8 categories (Native Forest, Reforestation, Water Bodies, Exposed Soil, Pasture, Corn, Sugar Cane and Soy) in the territory of Vale do Ivaí - PR through of remote sensing and machine learning algorithms, and to identify the patterns of soil transitions that occurred in the analyzed period through the use of intensity analysis methodology. To carry out the classifications of the studied territory, two models of machine learning were trained and tested, a shallow artificial neural network (SNN) and a deep artificial neural network, popularly known as deep learning (DNN). According to the results obtained, we can say that the SNN algorithm was the one that stood out the best, presenting the values of global accuracy and Kappa Index, 98.95% and 98.77% respectively. Through the application of the intensity analysis methodology, it was identified that the pasture category was the one that lost the most area, and that the sugar cane was the category that most took up the areas, in other words, sugar cane advanced intensely over the areas of pasture. We can conclude that this transition pattern occurred due to agriculture and the culture of sugar cane in particular, tending to expand over areas with low productivity, and / or with low yields
     
    URI
    https://hdl.handle.net/1884/100998
    Collections
    • Engenharia Agrícola (Jandaia do Sul) [24]

    DSpace software copyright © 2002-2022  LYRASIS
    Entre em contato | Deixe sua opinião
    Theme by 
    Atmire NV
     

     

    Navegar

    Todo o repositórioComunidades e ColeçõesPor data do documentoAutoresTítulosAssuntosTipoEsta coleçãoPor data do documentoAutoresTítulosAssuntosTipo

    Minha conta

    EntrarCadastro

    Estatística

    Ver as estatísticas de uso

    DSpace software copyright © 2002-2022  LYRASIS
    Entre em contato | Deixe sua opinião
    Theme by 
    Atmire NV