ect
but
] to
;.
ical
uch
ub-
lere
: a
ong
can
ra
and
uch
' of
the
uch
hic
as
can
ctor
hic
the
hic
. In
CG
d in
] to
to
of
ID
Jes:
the
into
the
| of
es,
tion
of
)hic
the
age
the
felt
ring
at
According to the above-mentioned requirements we
should be able to pinpoint the following types of
knowledges needs for implementation of ACG
(Buttrnfield, 1991): 1) structural knowledge;
2) geometrical knowledge; 3) procedural knowledge;
4) application requirements.
The first two types of knowledge can be used during the
first stage of CG. Here the four types of abstraction of a
GO into its digital representation are used: classification,
association, generalization and aggregation
(Buttrnfield, 1991). All these types of abstraction are well
described by the OOA. OO multi-detail GO model
supports such types of relations as generalization
(inheritance), aggregation (composition) and multiplicity
of association. Such characteristics can be applied for
multi-detail representation of a GO at grouping
(aggregation) of various types; progressive refinement of
subclasses in the form of hierarchy structures with
inheritance of characteristics and behavior from a
superior class; introduction of semantic multiple links
between objects of the same level.
Besides, as it is supposed that GO is composed of
geometrical subjects, such characteristics of the OOA as
structural OO and encapsulation are of utmost
importance. Such traits of the OOA support design of
complex objects and representation thereof or their
components at various stages of activities.
The set of GO classes has not been pre-fixed and a user
shall be able to extend it. And such properties of OOA as
overloading, polymorphism, behavioral and late binding
can enable this.
One of the important traits of OOA for support of multi-
detail representation of GOs is dynamic binding. This
characteristic of OOA enables real-time transfer of
changes performed at base level of the entity model, to
lower-detail levels that are used as applications.
At the second stage a symbol model of a GO utilizing
3rd and 4th knowledges, can be created on the basis of
the OO model of a geographical model, having such
traits as inheritance, association, aggregation,
encapsulation, overloading, polymorphism, etc. Here
OOSMDM is used for multi-scale analysis of data and
OO cartography modeling. As a result of analysis of
OOSMDM a user should be able to get a map of the
required scale.
A GO exists in the OOGDB based on the principle of
OOA identification. This principle means unique
properties of each object and existence of each object
apart from its notions.
During design of a CO of a certain level of details which
would comply to the prescribed scale, the message
mechanism of OOA is very important. Thanks to these
characteristics, a OOA object is able to communicate
with PC and other objects, and to perform all
computations needed for its scale representation and
solving of graphic and semantic conflicts between
adjacent objects during cartographic visualization.
The following chapters reflect structural MGIS schemes
and some of the groups of its classes. The links between
classes are based on the above OOA principles. This
overview of MGIS classes is illustrated by the following
notions — generalization, «—» aggregation,
— association (one to one) —> association (one
to many) (Rumbaugh, 1991).
THE OBJECT-ORIENTED MODEL OF MGIS SPATIAL
OBJECTS
The MGIS Spatial objects can be divided into abstract
and conceptual ones. This is application approach to the
classification of object's classes.
A geographic object is simulated in MGIS with the aid of
spatial objects of five hierarchical classes:
- inneror elementary geometric objects;
- Simple geometric objects;
- complex geometric objects;
- digital models of geographic objects (GO);
- cartographic objects.
The spatial objects of the first three classes belong to
abstract spatial objects, but acc. to their type, they
belong to concrete ones. The digital and cartographic
model of a geographic object is formed of the spatial
objects of the geometric classes.
Each object at a basic level is formed of two essences:
records and relations, the above-mentioned level being
no object-oriented level but, as a matter of fact, a
semantic network.
A record can have an arbitrary structure; in such a case
the DBMS has no idea about its content. The record can
change its dimensions. Anything can be stored in it,
including executed code, generalized line-tree
(Jones, 1991, van Oosterom,1995), R-tree, etc. To gain
access to the record, methods - procedures are available
(e.g. methods of implementing generalization
techniques, determination of generalization conflicts,
plotting of implicit Delaunay pyramid, etc.).
Record are joined by relations. A relation indicates two
records, and is characterized by a direction and type.
Each record has references to all its relations. Each
relation has a unique identifier. A type of relation also
represents a record. The latter can be modified, too. A
type controls records belonging to it.
A special record - procedure contains a set of operations
for a definite type. It is connected with this type by a
relation of a respective kind. While dealing with a record
belonging to a definite type the respective procedure is
recalled from the set of operations. The procedure works
at the contents of the records.
Inheritance is realized through types and relations. One
type is related to another in a hereditary way. The
system recalls operations: first for the derivative type,
then for the basic one. Inheritance can be plural.
The indexing of records is carried out by inserting a
special index record-type ensuring fast execution of an
inquiry.
The object can belong to several types, and change a
type. It must not necessarily belong to a definite type.
The conceptual data model is formed of abstract and
concrete objects and methods (permissible operations
with objects), and is object-oriented. Conceptual objects
are realized by writing the respective set of procedures
processing inquiries to records of the given type (e.g.
“inquiry about the value of the field or as to the possibility
447
of changing the value of the field).
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996
