DESIGNING AND PLANNING OF CLOSE-RANGE PHOTOGRAMMETRIC NETWORKS: IS AN EXPERT SYSTEM
APPROACH FEASIBLE ?
Bammeke A.A. and Baldwin R.A.
Department of Land Surveying, Polytechnic of East London
Longbridge Road, Dagenham, Essex, RM8 2AS. UK
Commission V
ABSTRACT
Designing and planning of close-range photogrammetric (CRP) networks require the solution of
a number of inter-related problems.
recording cameras, targeting,
instruments and data acquisition schemes.
Decisions have to be made concerning imaging geometry,
data-processing
algorithms, image-coordinate measuring
Some aspects of the decision-making are cognitive
in nature and are not suitable for a conventional algorithmic solution. These are therefore
not incorporated into existing network design packages.
solution to problems involving cognitive decisions.
Expert system technology offers a
This paper investigates the application
of expert system technology to CRP network design, and describes an experimental system which
has been applied to close range problems.
Some examples are presented which demonstrate how
the system facilitates the decision-making process.
KEY WORDS: Close-range, photogrammetry, network design, expert system.
1. INTRODUCTION
It is usual to perform network design as a
‘prelude to undertaking a close-range
photogrammetric (CRP) survey. Based upon an
initial choice of parameters,
photogrammetrists may simulate the results of
the initial configuration before proceeding to
carry out the later stages of photography,
measurement of image-coordinates, and
adjustment or data processing. Increased
automation has brought about a continuing
shift of emphasis to network design.
Network design involves the solution of a
number of inter-related problems. Decisions
have to be made concerning the choice of
imaging geometry, recording cameras,
targeting, data-processing algorithms, image-
coordinate measuring instruments and data
acquisition schemes. Some aspects of the
decision-making are cognitive in nature, that
is they can be made only on the basis of
knowledge gained from a combination of
practical experience with CRP measurements,
intuition and 'rules-of-thumb'. Cognitive
decision-making is not suitable for a
conventional algorithmic solution, and is
therefore not incorporated into existing
network design packages. Many workers, eg Chen
(1985) and Shortis and Hall (1989), have
emphasised the need to develop an interactive
computer package for handling network design.
A great deal of interest has arisen lately in
the application of expert system technology to
problems in which computer solutions were
previously inapplicable. This technology has
been employed in a variety of science and
engineering environments to solve problems
involving cognitive decision-making. An expert
system is yet to be developed for designing
CRP networks (Shortis & Fraser, 1991).
The aim of this paper is to demonstrate the
potential of expert system technology to the
planning and designing of CRP networks.
2. EXPERT SYSTEM TECHNOLOGY
2.1 Definition and structure of an expert
system
Expert systems are species of computer
software which use specialists' knowledge and
reasoning techniques to provide advice and
454
counselling, and to solve problems that would
normally require the expertise, abilities and
experience of human specialists. They assist
decision making and allow interactive consultation.
Expert systems differ in a number of respects from
conventional computer programs such as database
management systems (DBMS) or spread sheets. For
instance, in expert systems, :
(a) the bulk of a 'program' is made up of
statements of facts (or rules) rather
than control structures eg IF...THEN
is a relationship and not a control
structure,
(b) the physical order of rules are
irrelevant, since manipulation is not
done sequentially according to fixed
algorithms,
(c) answers can be provided not only to
the first order question (ie 'what?'),
but also to the second and third order
questions (ie 'how ?' and 'why ?')
Expert systems can be divided into two general
categories according to the task they perform
(Kretsch, 1988): those that design something in
order to solve a problem within some set of
constraints or guidelines, and those that perform
diagnosis (analysis). In either case the basic
structure is the same.
A typical expert system has four main components:
'knowledge-base', 'inference engine', 'working
memory', and 'user interface' (Fig.1). The
knowledge-base contains structured and codified
information about a specific problem area. In most
expert systems the knowledge-base is represented in
the form of rules. The 'inference engine' is a set
of computer programmes which constitute the central
problem-solving mechanism that controls and
coordinates the operation and reasoning of the
expert system (eg Ripple and Ulshoefer, 1987). It
is like the 'interpreter' or controller in
conventional programming (Sarjakoski, 1988). It
runs the program; matches rules with data;
determines which of the possible set of rules
and/or facts in the knowledge-base is to be applied
at each step, and when and how to use them for the
current consultation session. The 'working memory'
contains the description of the current state of
the problem-solving (Sarjakoski, 1988), and the
intermediate hypotheses; while the 'user interface'
controls how the user may communicate with the
system. A user can interact with an expert system
either by first suggesting a hypothesis, or by
first volunteering some data.
