A thorough investigation is reported on the qualitative modeling of geologi
c systems, focusing on the reconstruction of three-dimensional (3-D) profil
es from image data by means of spatial and temporal reasoning techniques.
A conceptual model of the relevant knowledge is proposed for both the domai
n elements and the inference processes. At the former level, we describe th
e objects in terms of geometric primitives and relations among them; at the
inference level, reconstruction is identified as a synthesis task, in whic
h a 3-D model of underground bodies results from assembling simpler compone
nts. The process is incremental and nonmonotonic, according to a basic Asse
mble-Validate-and-Debug cycle, underlying both low-level and high-level ste
ps. A formal (logical) model of the latter is proposed and worked out in de
tail. Concepts from topology and graph theory provide effective tools to de
fine representations and algorithms, and allow to address the intertwining
of spatial and temporal knowledge.
Some relevant reasoning steps are also regarded as constraint satisfaction
problems. We analyze the constraints, show that the related tasks can be so
lved with algorithms of polynomial complexity, and provide the appropriate
procedures. The practical feasibility of the model has been tested, and res
ults of the applications to realistic input data are discussed. We also dis
cuss solutions for embedding the modules into a man-machine interface for t
he intelligent support to the interpretation of data.