Exploiting spatial and wavelength information in dual-frequency SAR mapping of human settlements and their surrounding environment in the Amazons

This paper shows the first results of a partially supervised approach to dual-frequency SAR data classification in the Amazons. The target is not only to detect and extract human settlement extents and their features, but also to capture the surrounding environment with accuracy values comparable to manually digitized map. The results in the test area of Paragominas (Brazil) show that human settlements extents can be mapped to excellent levels, detecting critical areas in the manually digitized maps. Less accurate results may be achieved outside these settlements, and require a more complex and refined approach.