Classificação de imagens em sensoriamento remoto: desenvolvimento de uma ferramenta para identificação de áreas de incerteza

Abstract

Remote Sensing provides Earth data with consistency and periodicity, and can be understood as a practice of obtaining information about the Earth and surface of water with the images acquired from an aerial perspective, using electromagnetic radiation in one or more of the regions of the electromagnetic spectrum, reflected or emissioned of the Earth’s surface. However, images of Remote Sensing require through computational processing. Process an image is a method of successive transformations that generate groups detached from each other from the detection of patterns, allowing information to be obtained more easily. This work developed a tool in process of image segmentation, offering the choice of two grouping algorithms, the Fuzzy C-Means and the Fuzzy ckMeansImage, in which the second, brings as a big differential the option of selecting the α-cut parameter, determining the minimum degree of pertinence of each pixel to its respective cluster.The algorithms were applied in two images of the same area, located in the western border of the state of Rio Grande do Sul, at different times (Rio Uruguai at normal level and during flood season). The results are presented and discussed throughout the work.