72 
file the location has additional advantages since we can manage 
image data through internet. 
4.1 Using 2D-PIR for Content based metadata 
2D-PIR that proposed by M Nabil (1997) has an architecture of 
Image Retrieval System. The real image is considered as a big 
raster image file. Metadata of the image file is replaced by two 
type of file iconic image and a symbolic 2D-PIR 
transformation of the image. The iconic image is a small bitmap 
file that is stored in local storage as a preview of the original 
image data. A symbolic image is a spatial / topological 
information of the image file. This symbolic image is used as a 
representation of image content for image query. Using these 
two type of 'metadata', the user can query image data using two 
alternatives : selecting image based on image preview, or 
selecting image based on symbolic content. 
4.2 Storing spatial and temporal information 
The next step is building a database that contains image 
information, constructed for spatial and temporal information. 
There are many types of data that should be stored for every 
image to make a single record. An Object Oriented Database is 
recommended for this purpose. Some important objects that 
related to image information and should be stored in the 
database are : 
a. Topology 
A topology object is a directed graph that represents a 2D- 
PIR of image content. This spatial information is made by 
manual sketch using vector editor, or semi-automatic 
creation using Artificial Intelligence or edge detection. All 
topology files are stored in local storage for faster access. 
Topology file will be used to query image file based on 
image content. 
b. Image Preview 
An image preview is a small raster image in JPEG or GIF 
format, created from original image. An image preview file 
will be used as a sample file for image query using 
similarity of spectral information. This preview is also used 
as a preview output of image query. 
c. File Name 
A filename is a standard O/S file name. 
d. URL / Search path 
The URL / Search path is a pointer to file location. In local 
storage or LAN, a search path is a volume name, followed 
by a full-path directory name where the original image file 
stored, for example : G:\IMAGEVAVHRR\JAVA. If data is 
available through internet or intranet, a search path can be 
replaced by a URL, using an IP number or Internet Address, 
for example : 
http://leviathan.tamu.edu:70/1 s/slides/Corel 
e. Image Title 
An image title is a string, for example 'NDVI for West Java' 
f. Time Stamp 
A time stamp is a valid time of data acquisition. The user 
can refer to text-based metadata or header file to find this 
information. 
g. Image format 
Image format information is a string that describe what 
application should be used (or launched) to open this image, 
for example ER-Mapper, PCI or ERDAS. If image file 
format can be used for more than 1 application, this object 
notes the image file type, e.g : BIL , LAC. 
h. Coordinate 
Coordinate system includes all projection information if 
any. 
i. Sensor 
Sensor information contain sensor type and number of 
bands. 
4.3 Query Possibility 
From database structure proposed above, we can make image 
queries based on historical information, Image definition (title), 
spectral similarity or image feature that is stored on spatial 
information. Some example of query that is possible to be 
developed : 
Find image of East Java before 01 Jan 1988 
Find NDVI of Indonesia between 01 Jun 1995 and 31 Dec 
1995 
Find image that similar to image "BALI" 
Find image after 31 Dec 1997 that match this topology. 
5. Future Work 
Based on this conceptual design, development of this temporal 
data handling requires a computer language that supports Object 
Oriented programming and the ability to develop Object 
Oriented Database. Java and C++ Language would be the best 
language for this purpose. A Graphical User Interface (GUI) 
should be considered as an interface. For internet / intranet 
access, Java is easier implemented than C++. 
Some future applications related to this topic are : 
5.1 Image Query based on feature 
Data Query is widely used in relational database, known as 
Structured Query Language (SQL). In temporal database, data 
query is currently under development. There are many 
prototypes of a Temporal SQL. Similarly, Image Query Based 
on Content or Feature is a new issue for multimedia objects. 
5.2 Dynamic modelling and prediction using satellite 
imagery 
Considering the occurrence of time type information, we can 
perform dynamic modelling and prediction using satellite 
imagery. In current remote sensing and GIS application, It is 
only possible to perform spatial non-temporal modelling. 
For Satellite imagery that alfeady contains spatial and temporal 
information, dynamic modelling and prediction can be 
implemented using mathematical models. 
Some remote sensing satellites can capture one specific area 
everyday. We can develop much more dynamic models 
incorporating temporal data using daily information from 
satellite imagery.