Constructing a prior information base for river mapping from digital images and DEMs by an advanced image interpretation system

The purpose of this dissertation is to describe the construction of a prior information-base (interpretation of river characteristics) for river mapping from digital representations, such as remotely sensed digital images and DEMs, by an advanced image interpretation system. More reliable prior information availability in an advanced image interpretation system enables GIS and remote sensing facilities to locate rivers in an easier, more accurate and more straightforward way. In this study, the author proposes a prior information-base including some rules and facts for river mapping from the use of both remotely sensed multi-spectral images and DEMs. These rules not only allow water-related applications in both GIS and remote sensing to be more accurate, but also construct the information-base for river mapping by an advanced image interpretation system. These rules were constructed as a synthesis from searching the literature and experiments with both digital images and DEMs. These rules are employed in an advanced image interpretation system which requires (1) a prior information-base; (2) a working memory; (3) an inference module (Caelli and Bischof, 1997; Schenk and Zilberstein, 1990). A prior information-base is formed by a set of rules (qualitative or quantitative or relationships). A working memory has the basic function of holding features in the form of spatial data and their attribute values. These spatial and attribute data are used by interpreters to activate the rules. Inference module refers to software and hardware that connect the user's questions to the prior information-base and instruct the user (interpreter) about the process. In this dissertation, a prior information-based river mapping was implemented and the proposed rules were tested studying both images and DEMs in IDRISI, as well as RiverTools.