The aim of this study was to evaluate the performance of classifiers support vectors machine (SVM) and K-nearest neighbor (K-NN) for object-based image analysis using Sentinel-2 images. Três Pontas city in the southern region of Minas Gerais was used as a study area. Sentinel-2 image with a spatial resolution of 10 meters was obtained by merging and resampling 10 of all 13 bands. Based on prior knowledge of the landscape were defined 5 classes of use and land cover. The step of image processing occurred in ENVI 5.0. In segmentation step was applied to 10 values of "segment settings" that uses the algorithm "edge" and 60 for "merge setting" using the algorithm "full lambda schedule", respectively and targeting settings and unity. After that was collected train samples of all 5 predefine class. The classification was performed by SVM and K-NN algorithms. Both obtained satisfactory results in evaluation of accuracy with Kappa values for the SVM of 0.87 and 0.85 for K-NN. The results show that the object-based image analysis using Sentinel-2 images are robust and satisfying. The method allowed the correct classification of different vegetation types found in landscape. Furthermore, is recommended for preparation maps of land use and land cover that may assist the territorial planning.