Padrões binários locais na classificação de imagens de sensoriamento remoto

This paper investigates the use of texture descriptors based on Local Binary Pattern (LBP) for applications of automatic interpretation of remotely sensed (RS) images. After describing the use of LBP for applications in RS, the paper reports experiments involving land use and land cover classification respectively on Quickbird-2 and IKONOS-2 imagery. Different configurations for each texture representation approach have been considered in the analysis. In all experiments the LBP features were consistently superior to the traditional Grey-Level Co-occurrence Matrix (GLCM) based texture descriptor in terms of classification accuracy.Pages: 7651-765