%0 Conference Proceedings
%4 sid.inpe.br/marte2/2019/09.06.17.17
%2 sid.inpe.br/marte2/2019/09.06.17.17.53
%@isbn 978-85-17-00097-3
%T Mapping eucalyptus plantations and natural forest areas in Landsat-TM images using deep learning
%D 2019
%A Ferreira, Matheus Pinheiro,
%A Cué La Rosa, Laura Elena,
%A Happ, Patrick Nigri,
%A Theobald, Raissa Brand,
%A Feitosa, Raul Queroz,
%@affiliation Instituto Militar de Engenharia (IME)
%@affiliation Pontifícia Universidade Católica do Rio de Janeiro (PUC-Rio)
%@affiliation Pontifícia Universidade Católica do Rio de Janeiro (PUC-Rio)
%@affiliation Instituto Militar de Engenharia (IME)
%@affiliation Pontifícia Universidade Católica do Rio de Janeiro (PUC-Rio)
%@electronicmailaddress matheus@ime.eb.br
%@electronicmailaddress lauracue@ele-puc-rio.br
%@electronicmailaddress patrick@ele.puc-rio.br
%@electronicmailaddress
%@electronicmailaddress raul@ele.puc-rio.br
%E Gherardi, Douglas Francisco Marcolino,
%E Sanches, Ieda DelArco,
%E Aragão, Luiz Eduardo Oliveira e Cruz de,
%B Simpósio Brasileiro de Sensoriamento Remoto, 19 (SBSR)
%C Santos
%8 14-17 abril 2019
%I Instituto Nacional de Pesquisas Espaciais (INPE)
%J São José dos Campos
%P 2650-2653
%S Anais
%1 Instituto Nacional de Pesquisas Espaciais (INPE)
%K Convolutional Neural Networks, patchclassification, random forest, satellite images, tropical forests.
%X Automatic mapping of planted and natural forests using satellite images is a challenging task due to spectral similarity issues. In this work, we assessed the use of Convolutional Neural Networks (CNNs) to discriminate between natural forest areas and eucalyptus plantations in a Landsat-TM scene. First, we produced training and testing datasets with data from the MapBiomas project. Then, CNNs were trained with input patches of different sizes (55, 77, 9 9 and 11 11 pixels) to evaluate the influence of patch dimension in the classification accuracy. For comparison, pixel-wise and patch-classification were performed using the Random Forest (RF) algorithm. The best results were obtained using CNNs with 5 5 patches. In this scenario, the F-score was of 97.64% for natural forests and 95.49% for eucalyptus plantations. The classification errors reached 9.06% using RF and did not exceed 3% with CNNs.
%9 Floresta e outros tipos de vegetação
%@language pt
%3 97365.pdf