The objective of this work was to use object‑oriented classification in TM/Landsat‑5 images to characterize land use and land cover classes in the Araguaia region. The scene 223/068, acquired on 9/5/2010, was subjected to the following pre‑processing stages: radiometric, atmospheric, and geometric corrections. Two images were generated by the mathematical spectral bands normalized difference vegetation index (NDVI) and modified normalized difference water index (MNDWI), which were used in the classification process. For image segmentation, the scale parameters 250, 200, 150, 100, 50, the algorithms assign class and nearest neighbor, and the attributes of average, area, and border ratio were used. A confusion matrix was used to assess the accuracy of the classification, using the overall accuracy coefficient and the Kappa index of agreement. Overall accuracy for mapping was 83.3%, with Kappa coefficient of 0.72. The classification was done as to Cerrado physiognomies, anthropic and urban use of the land, water bodies, and sand banks. The mathematical spectral bands used are promising for delineating classes of the land cover in Araguaia.
digital image classification; Cerrado physiognomies; MNDWI; NDVI; image segmentation; remote sensing
