civil 
him 
n be 
)ase. 
soil 
this 
pouk 
n all 
bout 
were 
ural 
Tom 
tary 
lite 
for 
"the 
sing 
of 
| of 
and 
key 
age, 
rent 
sing 
tion 
vere 
vere 
'ode 
ase. 
ling 
It 1s 
1ds, 
ems 
ces. 
s of 
The search strategies used in developing the 
system were forward and backward search 
methods. ‘To match the method of search that 
an expert has in mind while conducting image 
interpretation, the main method of search used 
in this project was backward chaining. 
The primitive units of a class representation is 
a number of attributes and goals. The class 
was the landform and the attributes were 
topography, drainage, tone, gullies, vegetation, 
and laboratory tests (i.e. identification elements 
used in this research). Primitive facts are linked 
in conditional sentences by rules (IF-THEN- 
ELSE). The following is an example of typical 
rules in the system: 
IF The Topography OF The Image IS Flat table rocks 
AND The Drainage OF The Image IS Angular dendritic 
AND The Tone OF The Image IS Light banded AND The 
Gullies OF The Image IS Few to none AND The 
Vegetation OF The Image IS Barren THEN The Landform 
: = "The Landform is Sandstone" 
3.5 System Testing 
The system passed through two phases of 
evaluation. The first phase was a debugging or 
editing phase. It started as soon as the 
knowledge engineer began to program the 
system and continued until the system was 
completed. The second testing phase was an 
evaluation process of system performance as 
compared to certain reference (e.g. human 
expert). This stage was conducted when the 
system was operational and ready be delivered 
to the end users. 
4. SYSTEM IMPLEMENTATION 
The major components of this system includes: 
input data, inference mechanism, user interface, 
output and feed back modules. Figure 1 shows 
these components. 
The sets of input data were acquired by image 
interpretation. In this expert system one goal 
was assigned for identifying the expansive soil. 
The task of the inference mechanism while 
51 
conducting an application session may be 
viewed as intelligent tracking of a goal from a 
collection of classes, attributes and their 
method, rules, demons, displays and data bases. 
Class allows the developer to reference and 
control the functions, flags, and current states 
of an application. The goals of an agenda 
contain attributes that can be linked to method, 
display, rule and conclusions. 
4.1 Case Study For System Implementation 
Tabouk was selected as a sample area in Saudi 
Arabia. It undergoes severe problems due to 
expansive soil (Al-jawhra, 1989). Tabouk soil 
is expanded as soon as it has some water 
content. Its drainage system is so poor. It is 
impervious soil. Due to poor permeability and 
due to existence of minerals that help expansion 
of soil in Tabouk, poor civil engineering 
locations must be identified. Accordingly, the 
problem in this research was defined as 
follows: 
Tabouk contains expansive soils that require 
experts to define it in the real world and to 
inform people about its impact and poor 
suitability for civil engineering projects. 
For this study, Tabouk area (Figure 2) was 
divided into 4 regions. Two of the regions were 
used to obtain data necessary for the 
development of the system (see Figure 3). The 
other two regions were used to test the system. 
The major concepts explained so far for 
developing the system were executed on the 
two regions assigned for the development. 
That is, problem identification, data acquisition 
and data presentation were conducted using the 
acquired information from the two regions of 
the case study area. Samples of collected, 
processed, and prepared data for system 
development are presented in Table 1. 
Moreover, many images were processed for the 
study area. Sample of these images are seen in 
Figures 4 and 5. Also, laboratory tests of soil 
and field visits participated as important data 
for the development. 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996