A new approach to identify land use and land cover areas in Brazilian Amazon areas using neural networks and IR-MSS fraction images from CBERS satellite

This paper shows the classification obtained with an artificial neural network to map land cover areas in Brazilian Amazon region. The new approach is based on fraction images generated by linear spectral mixture modeling and used as input to the network. It identified with good accuracy the following classes: water, deforested areas, forests, and areas without predominant forest physiognomy (savannah and regeneration areas).