%0 Conference Proceedings
%4 sid.inpe.br/marte2/2017/10.27.13.10.36
%2 sid.inpe.br/marte2/2017/10.27.13.10.37
%@isbn 978-85-17-00088-1
%F 59282
%T Comparação no mapeamento da cultura de milho safrinha utilizando Machine Learning em imagens Landsat-8
%D 2017
%A Almeida, Luiz,
%A Johann, Jerry Adriani,
%A Richetti, Jonathan,
%A Nicolau, Rafaela Fernandes,
%A Richetti, Amanda Bordin,
%@electronicmailaddress almeidalz@hotmail.com
%E Gherardi, Douglas Francisco Marcolino,
%E Aragão, Luiz Eduardo Oliveira e Cruz de,
%B Simpósio Brasileiro de Sensoriamento Remoto, 18 (SBSR)
%C Santos
%8 28-31 maio 2017
%I Instituto Nacional de Pesquisas Espaciais (INPE)
%J São José dos Campos
%P 3455-3459
%S Anais
%1 Instituto Nacional de Pesquisas Espaciais (INPE)
%X The objective of this study was to compare the mapping of winter corn, using Machine Learning in Landsat-8 images in 2016 crop. For the images processing the software R 3.3.1 and ArcMap 10.0 were used. From a false-color RGB-564 composition of the Landsat-8 images 5 classes of soil use and cover (urban area, water bodies, forest, winter corn and exposed soil) were polygonised. These sampled areas served as training data for the models. The Random Forest and the Gamboost classification methods were applied. To perform the accuracy of each mask random points were generated for each classification and a being point-to-point verification was performed. For the Gamboost method the value of the adjustment parameter that allowed the best result was 150 iterations (Mstop). While Random Forest presented the best classification result when the number of predictors sampled in each node (Mtry) was equal to 2. The winter corn area identified in each model was about 75,290.58 ha for GB and 57,220.29 ha for RF, with Global Accuracy of 87.75% and 79.0%, respectively. In spite of the differences between the classifiers used, both methods are effective in mapping the studied culture. Moreover, both methods presented great agility to classify and to obtain area, aiding in the ergonomics of the processes.
%9 Classificação e mineração de dados
%@language pt
%3 59282.pdf