%0 Conference Proceedings
%4 sid.inpe.br/marte2/2017/10.27.16.17
%2 sid.inpe.br/marte2/2017/10.27.16.17.03
%@isbn 978-85-17-00088-1
%F 59365
%T Processamento de imagem SAR (Banda L) para detecção histórica de áreas florestais degradadas por incêndios recorrentes em Roraima
%D 2017
%A Xaud, Haron Abrahim Magalhães,
%A Santos, João Roberto dos,
%A Martins, Flora da Silva Ramos Vieira,
%A Xaud, Maristela Ramalho,
%@electronicmailaddress haron.xaud@embrapa.br
%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 7216-7223
%S Anais
%1 Instituto Nacional de Pesquisas Espaciais (INPE)
%X This paper aims to analyse the historical images of ALOS/PALSAR data (L Band) as an alternative to monitoring degradation of tropical forests affected by fires in the Northernmost Amazon Region. The sites of this study are located in the Apiaú Region, State of Roraima, Brazil. The study area was burned irregularly in 1998, 2003 and 2007 fires. The post-fire image (Jan.2008) was obtained in HH polarization. We ortoretified the PALSAR data and generated Amplitude and Intensity images. Additionally it were generated 13 textural data based on occurrence and co-occurrence matrix. Using Object-Based Image Analysis (OBIA) we segmented a 2007 Landsat TM image (as reference) to obtain objects that were described by 15 attributes derived from SAR images plus the standard deviation (SD) of each one, totalizing 30 attributes per object. We selected training and reference samples divided into 5 classes: (FN) unburned forests; (FQ1B) forests affected by 1 fire-low intensity; (FQ1A) forests affected by 1 fire-high intensity; (FQ2) forests affected by 2 fires; (FQ3) forests affected by 3 fires. We optimized the selection of PALSAR attributes to obtain the best separability among classes using a feature space optimization tool in OBIA based on Nearest Neighbor classifier. From the 30 attributes derived from PALSAR image, the results highlighted the best attributes (images) to detect degraded areas by recurrent fires; eight of them obtained from SD of textures and amplitude images.
%9 Degradação de florestas
%@language pt
%3 59365.pdf