The dynamics of the land-cover change, appears as the most important task at the present time, in which this activities have a direct influence on environmental resources available to society. The methods of the classification of remote sensing images are fundamental to understanding the dynamics of changing occupation of the territory, and are presented as a key tool for effective management of natural resources. This study aimed to define the best classifiers of images for the uses in the agricultural region of the center-east of Sao Paulo/Brazil, in the temporal cutouts 2001 and 2015, analyzing the ratings for each year, and between the two different years, using of error matrix and Kappa index. Was used images of Landsat (satellites 7 and 8), instruments ETM+ and OLI respectively, and processed in ENVI and ArcGIS. It was concluded that the classification supervised by distance of Mahalanobis should be used with caution in the event of clayey soils with high humidity, because the spectral signature of water is similar of soil wet, for this method of classification. In the analysis for the first year we obtained an overall accuracy of 85%, which is an good indicator of accuracy of the classificators selected. In the comparative analysis between the years under review, the overall accuracy was 26.3% and the Kappa index of 0.13, thus indicating that there was a significant change in land-cover. It is emphasized wich to carry out the land use classification, it is necessary to use more than one classifier.