TAX: Prototype Expert System for Terrain Analysis

Abstract

Terrain analysis is a time‐consuming, costly, and labor‐intensive process requiring special skills and training. Furthermore, an enormous amount of remotely sensed data is routinely generated by satellite and airborne sensors which can be used for terrain analysis. Thus, there is an urgent need for an automated approach to analyzing these data and model human reasoning. A rule‐based expert system methodology has been developed and the Terrain Analysis Expert (TAX) has been implemented for modeling interpretation logic involved in identifying landforms from aerial images. Knowledge about the geographic location of the image was used to arrive at hypotheses about the landform of the site manifested on the aerial image. These hypotheses were then established or rejected based on the degree of match between the hypothesized landform's pattern elements and those of the site. The site was declared to be the landform with which it had the best match. The pattern elements of the site were obtained interactively from the analyst. A probabilistic method was designed for handling uncertainties in the observed pattern element values and their role in the identification of landforms. The results indicated that a rule‐based expert system is appropriate for representing image interpretation logic involved in terrain analysis.