ISPRS, Vol.34, Part 2W2, "Dynamic and Multi-Dimensional GIS”, Bangkok, May 23-25, 2001
inheritance, encapsulation, congregation, etc.
Objected oriented geographic database need maintain the
integrality and consistency of storage and expression. For
example, a route has only one geometrical identification and one
geographic identification. The data organization of geographic
data includes organization and management of spatial data,
attribute data, image data, etc. The key technology of data
organization is to express and store spatial data and attribute
data integration.
1) spatial objects mainly divided into five categories:
• Point: It includes single point (e.g. buoy, navigation mark,
etc.), vector point (e.g. beacon), etc.
• Line: Line is interconnected points or arcs, they may netlike,
but must interconnected, such as fiber cable, etc.
• Region: Region was made up of one or many closed and
disjointed polygons and rings, it can include multi-island.
For Example, anchorage ground, restricted navigation zone
are all region.
• Complicated object: The object was made up of two or
more basic objects (point, line, region) or other complicated
objects. For example, school can act as a complicated
object,. It includes roads, houses, water tower and some
other basic geographic elements.
• Annotate. Annotate is used to describe place name,
features of geographic objects, etc. Because the model
need all terrain features should be stored integration, each
integrated terrain feature only has single geometric object
identification no matter how big the terrain feature is. The
spatial data of terrain feature should be stored integration.
But the segregation of integration of terrain features should
depend the requests of users and applications completely.
2) Attributes: Attributes of all spatial object should be managed in
project unitedly by RDBMS. Attribute data were stored in tables of
RDBMS. Referring to standard objects, users classify them
unitedly, and appoint single classification code (identification) of
terrain features. For example, we classify them into control point,
road, habitation, water system, vegetation, etc. in different tiers.
Any integrated terrain feature only corresponds to one record of
attribute data. Single object identification (OID) connects spatial
object and the corresponding attribute record, which was
described in figure 1. In order to organize and express spatial
geographic objects efficiently, we extract objects hierarchically
according to the size of geographic objects, and arrange,
delaminate and build spatial index for these objects.
The system architecture of object oriented chart database. The
B/S map database is the production of combination of database
technology and computer network technology. It is a computer
system to manage distributed map database. A distributed map
database is a set that was composed by many logic related
databases distributed in network. Each node of this network has
independent processing ability to implement local application. At
the same time, each node can implement global application
through network system. Map database server is mainly used to
manage SQL requests from clients, and return results to clients.
In practice, database server achieves all data operations, spatial
data organizations, concurrency control, security auditing, and
system management in C/S database. Client mainly presides
send user’s SQL requests to server, and processes and
expresses the results. In a typical application structure, client
usually deals with the application’s expressive logic. B/S map
database is an organic association of member map database
located in nodes of network. The management of this association
is B/S map database system.
Electronic Chart evolved from file-based application to spatial
geographic extended RDBMS, and aim to a integrated object
relational system. Therefore, we can use spatial geographic
technology to process file, time series data, image and audio, and
some other standard, abstract data type seamlessly. The problem,
which must be resolved by RDBMS, can use more direct and
accurate approach (e.g. SQL) to resolve.
Traditionally, there are two kinds of spatial data index: The first is
to build a spatial index out of the server and stored in BLOB, and
you can use it or search BLOB when you use it. However, no
matter how fine this index’s structure is, the maintenance and
searching include external parts, which will lead to more I/O
overhead and coexist problem, to make system slower and
inefficient. Second method is spatial segmentation. This method
divides data into cells according to stated grid and other layered
architecture (e.g. quad tree). Each cell was assigned a number,
and each spatial object relates to the number of the same cell.
But space is not in linear order. The irregularity of many spatial
features needs intricate searching and the check for many
potential errors. It’s also a slower and inefficient method.
Comparing to the methods we discuss above, R-tree is a
high-powered, multidimensional access approach. This approach
is built in kernel of database and cooperates with extensional
data type to manage spatial geographic data properly. Being
different from standard index. R-tree does not divide the space
into integrated overlay made up of non-overlap and adjoining
cells. Adversely, it automatically represents each object with "side
box” that is decided by spatial figure. These “side box” may
overlap, and do not fill all space. We should not know the spatial
dimension of data in advance.
A spatial data engine (SDE) layer or data class is geographic
feature classes that have the same attribute class. In order to
search conveniently, SDE uses DBMS attribute column index.
SDE index all features in a layer, to realize rapid spatial searching,
and provides high-powered searching for oversize database. The
unique function of spatial index is to provide special two
dimensions index of spatial features, and to realize a logical and
nonphysical hierarchy. The efficient supporting of geographic
data not only needs point, line and region objects, but also needs
a set of steady spatial object models that can express their
features clearly. SDE can describe linear features, for instance,
an uncrossed road, a road crossed at termination, a self-crossed
rode, and multi-roads that crossed at end-point to form a network.
Similarly, we can describe region features by single polygon,
hollow polygon, and poly-polygons.
Attribute also cannot be directly indexed into regional piece
object, but indexed into a concrete place in this regional piece.
Due to each accessing of polygons need be included into
reconstruction of multi-accessing, the feature we talk above may
become a default. The spatial object model of SDE stores each
regional piece into an integrated polygon. Thus, we can find
entire object by only one accessing.
In SDE, geographic features include spatial objects (e.g. point,
line or region) and relevant attributes described by these objects.
In RDBMS, data were stored in the tables made up of rows and
columns. The crossed cell of row and column names field, and
the data stored in field names value. Row represents a specific
event, distance or geographic feature, and column includes the
attribute of this feature. There are many types in attributes, which
make geographic features become a value type stored in a
column. SQL provides a interface of rational table, which enable
user can select rows according to the number in the field. SQL
statement is very flexible to program any basic query request.
Through the key, user can query the required data from tables. A
key (it may be made up of one or more columns) represents the
table uniquely. A column or several columns can be copied to
another table, named external key. Main key and external key link
two tables together, which can reduce the redundancy of the
database. This technology is called database standardization.
We can satisfy the request and return the accurate geographic
data through adding more tables and using the relations among
main keys and external keys. The query results are a series of
rows that satisfy the corresponding SQL statements. These SQL
statements are called cursor. The application can index large
numbers of active cursors, and query every independent row
from a cursor.
In object oriented DBMS, SDE can make IDS store and manage
complicated geographic data. SDE does not change existing
database or affect current application. It only adds a “Shape”
column into existing table, and provides software to manage and
access geometrical features of this column. SDE use R-tree to
