Mining patterns of change in remote sensing images

The National Institute for Space Research (INPE) holds more than 130 terabytes of image data, which cover 30 years of remote sensing activities. The availability of such huge image set demands appropriate techniques to explore it. However, we have a limited capacity to extract information from these databases, once the main problem of information extraction from remote sensing images is to detect patterns of change. This thesis presents a proposal to extract patterns of change from remote sensing images through concepts of digital image processing, data mining and landscape ecology. Taking into account the social and environmental problems caused by the fast Amazon deforestation, this work supplies and evaluates resources to assist the comprehension of land use and cover change processes, as well decision making procedures related to this domain. The developed methodology was applied in remote sensing data, through a software prototype, to identify and analyze the deforestation processes in Amazon areas.