A blackboard infrastructure for object-based image interpretation
Overview
The project concerns a blackboard infrastructure for the development of
object-based image interpretation applications. The infrastructure is based on
an abstract definition of important concepts in image interpretation: data
objects, relationships, algorithms, strategies, and models. Building
applications with the environment is effectuated by specializing the abstract
concepts. Using abstract concepts as a basis guarantees a high level of
extensibility, flexibility, and re-usability.
The structure of the implementation of a sample application in the blackboard
infrastructure is depicted below.
To demonstrate the use of concepts the infrastructure a specialization of the
infrastructure for the domain of industrial object recognition has been
implemented. The specializations include implementation of basic data objects
often used in object recognition (lines, arcs, corners), relationships between
the data objects (adjoined and aligned), algorithms to compute the data objects
and their relationships, and a strategy to map the model onto the data objects
(based on interpretation tree search). Furthermore, a graphical user interface
has been build to facilitate specification of models of objects to look for.
The interpretation process and final result are also presented in a graphical
way (see below for an example).
People
References
- D. Koelma and
A. Smeulders.
A blackboard
infrastructure for object-based image interpretation.
In E. Backer, editor, Proceedings Computing Science in the
Netherlands, pages 136-147, 1994.
(Gzipped PostScript, 12 pages, 76581 bytes)
(PDF, 166132 bytes)
- R. Van Balen, D. Koelma,
T. K. Ten Kate, B. Mosterd, and A. W. M. Smeulders.
ScilImage: a
multi-layered environment for use and development of image processing
software.
In H. I. Christensen and J. L. Crowley, editors, Experimental
environments for computer vision & image processing, pages 107-126.
World Scientific Publishing Co. Pte. Ltd., 1994.
(Gzipped PostScript, 19 pages, 190612 bytes)
(PDF, 286216 bytes)
- D. Koelma and
A. Smeulders.
An image
processing library based on abstract image data-types in C++.
In Proceedings of the 8th International Conference on Image Analysis and
Processing, number 974 in Lecture Notes in Computer Science,
pages 97-102. Spinger-Verlag, 1995.
(Gzipped PostScript, 6 pages, 47095 bytes)
(PDF, 126986 bytes)
- D. Koelma.
A software
environment for image interpretation.
PhD thesis, University of Amsterdam, 1996.
(Gzipped PostScript, 175 pages, 836291 bytes)
(PDF, 1172499 bytes)
Sponsors
This research was sponsored by STW project AWI92-1691.