of 
vn 
tic 
od 
al 
es 
nt 
at 
us 
ot 
ely 
he 
ed 
tic 
his 
the 
ind 
SO, 
the 
tial 
for 
ins 
ss) 
to 
ally 
of 
nal 
the 
rea 
ory 
locations near urban environments. On this 
data set unsupervised classification is applied. 
This method results in ten spectrally different 
classes i.e. classes are scaled according to 
the level of emitting thermal energy. After that 
information category of the particular class is 
defined, and classes with stronger thermal 
radiation are separated (shown in different 
shades of yellow, Figure 1). 
These information about thermal radiation are 
used as a basis for the next set of data. The 
purpose of the new data set is a more precise 
definition of the area of interest. This second 
type of data for classification uses TM bands 
5,7 and previous band combination. We used 
this band combination because the warm 
objects have emission peaks that fall between 
0.5um and 9.5um and thus can be detected in 
MidInfrared TM Bands 5 and 7. Bands 5 and 7 
are useful for pinpointing small, warm targets 
because of their greater spatial resolution. 
Thermal infrared Band 6 is designed to 
measure surface temperatures. The unsuper- 
vised classification is used for this data 
combination (Figure 2). The shade of the 
yellow colour represents the warmer surfaces. 
Results given by the method are checked on 
the map. It is obvious that urban areas are 
specific for their increased thermal emitting. 
Comparing with map 1:25000 it is easy to 
eliminate classes related with urban objects 
and areas. The rest of the locations compared 
with the SPOT (p) images gives with large 
probability possible locations of garbage repo- 
sitory. 
The same method is applied on the variety of 
areas near urban regions for, which it is logical 
to assume that they are garbage repositories. 
The results were validated on the field. 
Polygons of the different garbage repositories 
define the minimal area cca 12000 m^ for 
which this method is still valid. So the method 
is correct for discovering larger “wild” 
repositories and for the controlling of the 
known repositories growth based on the 
satellite images taken in various times (Figure 
3). 
DATABASE DESIGN 
Project database was developed with 
ARC/INFO (rev. 6.1.1). Graphical data is 
modelled with line features and stored inside 
ARC/INFO line coverage. Attribute data 
describing garbage repositories is stored 
within two tables. One of them is arc attribute 
table (AAT) of the line coverage, and the other 
is standalone table called GARBAGE_T. 
Figures 5 and 6 show these tables with 
names, types and descriptions of the fields. 
This database design enables the user to track 
and store information about garbage 
repository changes during the time. First table 
(AAT) keeps data about each measurement of 
the garbage repository. GARBAGE_T table, on 
the other hand, keeps data that is time 
irrelevant and same for every measurement 
performed (i.e. commonly used name for the 
garbage repository or its geographic location) 
for each garbage repository. Each 
measurement (for all garbage repositories 
examined) has its own row in the coverage 
AAT table. Rows in AAT table that have the 
same value for the "cov-id" field represent the 
same garbage repository. Row representing 
the last measurement for one garbage 
repository has the value of the “validity” field 
set to *T" (true), and all other rows for that 
repository has this field value set to "F" (false). 
GARBAGE T table contains one record for 
each garbage repository examined. These two 
tables are related to each other with the field 
“cov-id”. 
The reason why the graphical data is modelled 
with lines and not with polygons is because 
this revision of ARC/INFO (rev. 6.1.1.) doesn't 
have support for overlapping polygons, but 
perimeters and areas data of garbage 
repositories are still required. Because of that 
some small programming was done to 
perform  digitalisation of tracked garbage 
repositories as polygons (in the separate 
temporary coverage), and after the necessary 
calculation, data is imported as lines in the 
main coverage. 
Capabilities of such designed GIS database 
besides digitising and storing new information 
of garbage repositories include search and 
display of data and measurement of garbage 
repository changes in the course of time. By 
that measurement the user can estimate 
growth of the garbage repository and its 
status, whether it is in every day use (active) 
or not (abandoned). 
CONCLUSION 
The experimental method is defined as a 
reliable way to detect garbage repositories on 
the available satellite images (Landsat TM and 
SPOT panchromatic). Also, a database design 
adequate for tracking of the garbage repository 
changes in the course of time is suggested. 
The method can as well be applied to detect 
wide uncontrolled repository in order to 
supervise the environmental polluting. Of 
course, analysis of the different satellite 
images with better resolution and radar 
images can ensure information of the other 
garbage repository caracteristics, and enlarge 
probability of smaller garbage repository 
detection. 
Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 
237