ClasSIS: uma metodologia para classificação supervisionada de imagens de satélite em áreas de assentamento localizados na Amazônia

Abstract

Amazon has one of the most complex and diversified biome of the planet. Its environmental preservation has an impact in the global scenario. However, besides the environmental factors, the complexity of the region involves other different aspects such as social, economical and cultural. In fact, they are intrinsically interrelated in such a way, that, for example, cultural features may affect land use/land cover. Moreover, depending on the scale of such changes, there may be consequences on the planet. Due to the criticality involving the region, several actions of governments, organizations and social movements and the international community, has been trying to rationalize the land use/land cover, in order to create a sustainable exploitation of available natural resources. An important government program is based on the creation of settlements, with respective support, with regard to certain infrastructure financing, machinery, seeds and seedlings, technical assistance, etc. However, despite the efforts made, the solution is extremely complex, given the extensive correlation of factors to be assessed and combined in search of success and improvement of such programs. So, because of all complexity involved, it is of paramount importance that there are methodologies that contemplate the complexity and inherent interdisciplinarity. Intending to have a closer look at this whole issue, an Amazonian set of institutions drew up and approved the project Skills Development and Training of Human Resources for Degraded Areas Recovery in Settlement Projects in Amazonian Areas – Public Notice Nº047 / 2012 - CAPES/Pró-Amazônia Programme. This work presents an innovative methodology for land use/land cover analysis, in projects of settlements located in the Amazon, as part of the above mentioned. Such methodology comprises two modules (image processing, and classification and patterns extraction), each module is subdivided into steps to performing their primary functions. The methodology proposed in this work, which has a high degree of generalization, aims to categorize the land use/land cover, through the association of each pixel to a predefined thematic class. In order to validate the proposed strategy, case studies are carried out in settlements located at the Southeast region of the Stat eof Pará, in the Brazilian Amazon.