Land cover mapping of Itatiaia National Park with OLI/LANDSAT 8 and InterIMAGE System
ISBN 978-85-88783-11-9
Authors
1Sousa, G.M.; 2Maia, L.H.; 3Fernandes, M.C.; 4Costa, G.A.O.P.
1FEDERAL RURAL UNIVERSITY OF RIO DE JANEIRO Email: gustavobond@gmail.com
2FEDERAL RURAL UNIVERSITY OF RIO DE JANEIRO Email: leonardo.rio@globomail.com
3FEDERAL UNIVERSITY OF RIO DE JANEIRO Email: manoel.fernandes@ufrj.br
4PONTIFICAL CATHOLIC UNIVERSITY OF RIO DE JANEIRO Email: gilson@ele.puc-rio.br
Abstract
Established in 1937, with an approximate total of 30,000 ha, Itatiaia National Park is the first Conservation Unit of Brazil. This Park is located in an Atlantic Forest reminiscent area, between the states of Rio de Janeiro and Minas Gerais, and faces serious problems related to the large number of burned areas of criminal origin combined with the lack of rain during the winter season (from June to September). The monitoring of this area with the help of thematic maps can be shown as an important resource for planning and decision making by their managers against the sprawl of wildfires. With advances in remote sensing technologies and softwares capables of interpreting the images in an automated way, the production of land cover maps can become simplified. However, high costs attributed to these software license and the images generated by satellites provided by some distributors can restrict access to these technologies. Thus, this paper aims to show an alternative methodology that allows the use of remote sensing through InterIMAGE open source software, developed by the Computer Vision Lab - LVC/PUC-Rio together with Image Processing Division - DPI/INPE, and OLI/LANDSAT 8 images traded at free costs by NASA. The image was atmospherically corrected before the computation of the classification with AtmCor4OLI software. From the construction of a semantic net modeled in an knowledge based system, proved the viability of producing thematic maps concerning the use and land cover, which enable the analysis of several components of the landscape as forest fragments. The methodology used as operators Baatz Segmenter, Shapefile Import, C4.5 Supervised Classification, Urban Focus and NDVI Segmenter and the results proved satisfatory to the global accuracy and kappa index in the identification of classes of vegetated areas (forest and grass fields) shadow area urban, rock outcrop, water and cloud.
Keywords
Classification; Geoecological Cartography; Knowledge model