nosmo@csilvax.ucsb.edu (Vince Kraemer) (05/19/88)
I am currently looking for research in computer vision that is
using a logic programming approach. So far, the search has been
less than successful. If you have any references to research along
these lines, I would be interested in hearing about it.
Thanks all,
Vince Kraemer
-------------------------------------------------------------------------------
Well, we got dis-claimers and dat-claimers, which claimer would you prefer ??
-------------------------------------------------------------------------------harwood@cvl.umd.edu (David Harwood) (05/19/88)
In article <632@hub.ucsb.edu> nosmo@csilvax.ucsb.edu (Vince Kraemer) writes: >I am currently looking for research in computer vision that is >using a logic programming approach. So far, the search has been >less than successful. If you have any references to research along >these lines, I would be interested in hearing about it. > >Thanks all, > Vince Kraemer >------------------------------------------------------------------------------- >Well, we got dis-claimers and dat-claimers, which claimer would you prefer ?? >------------------------------------------------------------------------------- I developed a large expert vision system for interpreting aerial photos, using Prolog to specify types and relations of objects in images, also to specify a metatheory (a sort of constructive model theory) for how to find a "large" interpretation of this image theory by suitably restrictive topological search in analyzed image structures. Image predicates and search procedures are evaluated by Lisp or C programs. There is a report in the April 88 Proceedings of the Image Understanding Workshop, at Cambridge, Mass., of application to interpretation of aerial photos of suburban neighborhoods. More recently I have been designing Prolog/C vision systems for biological research (mapping 3D neuronal branching structures visible by confocal light microscopy). [I am quitting military-funded vision research for conscientious reasons, so I haven't directly followed up the earlier work, which so far as I know is the first large logic-based system for image interpretation (considered as finding interpretations for Prolog "image theories.") I'm also aware of a 3d "shape from perspective" application published in CVGIP by a student of Haralick and Shapiro, in 1986 I believe. Another relevant report on potential use of logic programming for computer vision is "The Logic of Depiction," by R. Reiter and A. Mackworth, U. Toronto. I'm not familiar with their recent progress since this 1987 TR, which does not report actual design of systems, algorithms, or implementation, but which abstractly considers intrepretation of logical theories of images. David Harwood
mack@bay.cs.ubc.ca (Alan Mackworth) (05/21/88)
I responded privately on this query but since it has come up it's worth following
up. Ray and I have produced a logical framework for depiction and image interpretation
described in:
R. Reiter & A.K. Mackworth, "The Logic of Depiction"
RBCV-TR-87-18 Dept. of Computer Science, Univ. of Toronto, Toronto, ON, Canada,
also TR-87-24, UBC Dept. of Computer Science, Vancouver, B.C., Canada, 1987.
Abstract
We propose a theory of depiction and interpretation
that formalizes image domain knowledge, scene domain
knowledge and the depiction mapping between the image and
scene domains. This theory is illustrated by specifying
some general knowledge about maps, geographic objects and
their depiction relationships in first order logic with
equality.
An interpretation of an image is defined to be a logi-
cal model of the general knowledge and a description of that
image. For the simple map world we show how the task level
specification may be refined to a provably correct implemen-
tation by invoking model preserving transformations on the
logical representation. In addition, we sketch logical
treatments for querying an image, incorporating contingent
scene knowledge into the interpretation process, occlusion,
ambiguous image descriptions, and composition.
This approach provides a formal framework for analyzing
existing systems such as Mapsee, and for understanding the
use of constraint satisfaction techniques. It also can be
used to design and implement vision and graphics systems
that are correct with respect to the task and algorithm lev-
els.
We transform the specification from FOL+equality to propositional
form so our model theoretic approach has simply an NP-hard problem
(SAT) to deal with. The transformation can be seen as producing
a Constraint Satisfaction Problem. The known algorithms for CSP's
can be seen as approximation algorithms for SAT, in this case.
The natural implementation language for our framework is Prolog.
Alan Mackworth