Supported mapping with multi sensor images through strategy focused on customization and integration of generalized classes by GEOBIA

Abstract

ABSTRACT:Rio de Janeiro presents itself as a land of contrasts, with the second largest metropolitan area in Brazil and a total area of 43,778 km 2 . Its landscape consists of remnants of the Atlantic Forest and environments with different levels of anthropic impacts. The complexity of its territory combines natural and anthropic covers related to each other at different intensity levels, being a challenge in defining mapping techniques of land use and land cover with greater spatial and temporal detail. Supported by the Secretary of State for Environment, this initiative accepted the challenge of developing a methodology for mapping about 45% of the state in 1:25,000. This mapping aims to support decision making regarding the land use planning and the monitoring of deforestation actions. Using GEOBIA techniques and images of different resolutions, we structured a methodology for the classification of four macro-classes very different spectrally (Natural Forested Areas, Natural Non-Forested Areas, Anthropic Agropastoral Areas, Anthropic Non-Agropastoral Areas), identifying objects on the ground from 0.5ha, to meet demands from the Rural Environmental Registry (CAR, in portuguese) and the forest monitoring. The mapping was divided into Working Groups (WG) that customized solutions through Process Trees in eCognition environment. This choice is based on the need to minimize inconsistencies in the interpretive process, improving the level of specialization in short time. Thus, each WG is responsible for the classification of a set of functionally related classes. The integration process of the classes in a single mapping was also supported by GEOBIA. It is believed that the presentation of this strategic view may contribute to challenges of similar mapping, allowing the achievement of cartographic goals in short term.