Agrarian Reform in Brazil: Object-oriented classification and accumulated reflectance in monitoring and enforcement of land

Abstract

The use of remote sensing and geoprocessing techniques can optimize time, reduce costs and qualify the monitoring activities of land use and occupation by the National Institute of Colonization and Agrarian Reform (INCRA) in rural settlements. This work was intended to apply and discuss some of these techniques: the object-oriented classification of images of the Sentinel-2 satellite, using accumulated reflectance and data mining. The study was conducted in the areas of Colonia I and Colônia II rural settlements, in the municipality of Padre Bernardo / GO. In this study only free software were used: QGIS in the pre-processing of Sentinel-2 images, vector editions and preparation of the final layout of the classification map; InterImage in the segmentation of the images, definition and export of the attributes of each segment that based the classification and; Weka for the data mining of the chosen attributes, identifying the thresholds of the different classes to be mapped. The classifications with the accumulated reflectance technique improve the accuracy (global accuracy and Kappa index) as a whole, but only occurred in the classes related natural vegetation unlike the classes Cultures and Forestry. The programs used performed satisfactorily all the procedures proposed to achieve the objectives of the work demonstrating their applicability in the monitoring of the settlement projects.