%0 Conference Proceedings
%4 sid.inpe.br/mtc-m19/2011/02.08.11.07
%2 sid.inpe.br/mtc-m19/2011/02.08.11.07.44
%@issn 1682-1777
%T Object-based image analysis and data mining for mapping sugarcane with landsat imagery in brazil
%D 2010
%A Formaggio, A. R.,
%A Vieira, M. A.,
%A Rennó, C. D.,
%A Aguiar, D. A.,
%A Mello, M. P.,
%@affiliation Instituto Nacional de Pesquisas Espaciais (INPE)
%@affiliation
%@affiliation Instituto Nacional de Pesquisas Espaciais (INPE)
%E Coillie, E. A. Addink and F. M. B. Van,
%B GEOBIA 2010 Geographic Object-Based Image Analysis.
%C Ghent, Belgium
%8 29 June - 2 July
%I ISPRS Working Groups
%V 38-4/C7
%S Proceedings
%K Sugarcane mapping, Artificial Intelligence, Object-based Image Analysis, Data Mining, Landsat images.
%X Mapping of sugarcane planted area is an important information for decision making, mainly when the search for alternatives to mitigate greenhouse gas emissions has indicated the use of biofuels as a viable option. Thus, the aim of this research was to develop a methodology in order to automate the sugarcane mapping task when remote sensing data are used. Thus the integration of two major approaches of Artificial Intelligence, Object-Based Image Analysis (OBIA) and Data Mining (DM), were tested in a study area located in São Paulo state, which is well representative of the agriculture of large regions of Brazil and other countries. OBIA was used to emulate the interpreter knowledge in the process of sugarcane mapping, and DM techniques were employed for automatic generation of knowledge model. A time series of four Landsat images was acquired in order to represent the wide variability of the patterns during sugarcane crop season. Definiens Developer® multiresolution segmentation algorithm produced the objects and properly trained decision tree applied to the Landsat data for the generation of the thematic map with sugarcane as the main class of interest. An overall accuracy of 94% (Kappa = 0,87) was obtained, showing that OBIA and DM are very efficient and promising in the direction of automating the sugarcane classification process with Landsat multitemporal time series.
%@language en
%3 Formaggio_Full paper.pdf