ABSTRACT

Traditionally, the cartographic framework is based on the visual analysis of aerial photographs (Herold et al., 2003). More recently, high spatial resolution satellite images have also been used for photointerpretation, given the recent improvements of the spaceborne digital sensors. The success of the interpretation of images varies with the interpreter’s skill and experience, the nature of the objects, the analyzed phenomena and the quality of the images being used (Campbell, 2002). Visual interpretation is a method widely used for production of thematic cartography, not only because it achieves good results, but also because an alternative method is not always available. However, visual interpretation has some associated drawbacks. On one hand, the number of distinguishable grey levels for the human eye (approximately 16) is considerably smaller than the range captured by digital sensors. Similarly, the human eye can only compare three bands simultaneously (in a RGB color composite). But probably the prime disadvantage is that photo-interpretation is based on subjective assumptions. Green and Hartley (2000) found that the subjectivity in placing the boundary between elements

which gradually tend to each other is the factor which contributes the most to the positional error on a thematic map. This inconsistency may also create problems in map updating, even if made by the same person (Ahlcrona, 1995), rendering the conclusions based on the analysis of these maps, to be unreliable. In addition to the intrinsic characteristics of human recognition, the entire cartographic framework based on photo-interpretation requires time and allocation of resources.