TOMaTEX |
|
|
Introduction and Highlights on the Function of each Subsystem
|
The main goal of the tomato expert system, is to provide
its users with disorder diagnosis and recommendations about how to treat
these disorders. This task is complicated by the fact that the tomato crop is
cultivated in different environments,
e. g. under plastic tunnels, open fields, and low tunnels.
|
Functional Specification
|
The current version of the system (TOMaTEX ver. 3), provides recommendations concerning
the following agricultural activities:-
|
Diagnosis subsystem |
purpose
There are two purposes for this subsystem:-
- It concludes the causes of user complaint.
- It verifies the user assumption .
Output
The output of this subsystem is as follows:-
- If there are user complaints, the output is the causes (disorders)
of these complaints. Each disorder must have a certainty factor. The certainty
factor of the confirmed disorders is either "likely" or "most
likely".
- If the user knows the cause of the abnormal observations, and he wants
treatment for the assumed disorders, this subsystem must verify the user's
assumption.
|
Treatment subsystem
|
purpose
The purpose of this subsystem is to advice the user about the treatment
operation of the infected plant.
Output
The output of this subsystem is the treatment schedule. The output
includes a complete specification about the treatment operation: disorder
name, material name, material quantity, mode of entry, method of application,
the tool used in the treatment operation, application time, and advice.
The treatment subsystem takes into account the seriousness of the disorders,
and so it orders the treatment operations according to their seriousness.
Back to the beginning of this page
|
platforms and methodology
|
The Design Methodology and System Implementation :
What was required, was to build an expert system that is capable of performing
the diagnosis and treatment operations and that has the ability of employing images
to enhance its performance and interface with the ordinary user.
a panel of human domain experts were consulted. Knowledge elicitation sessions
for both the diagnosis and treatment subsystems were conducted for a total of about 90 hours.
plant disorders were classified into seven classes:
diseases, nutrition deficiency, environmental, insects, mites, bacteria,
and nematode. The diseases class was divided into two sub-classes, fungal,
and viral.
The following is the methodology that has been used in designing the
system :
- Dependency Networks were used as a means for communication
with the domain experts in order to acquire domain knowledge.
- KaDS was used for the representation of the inference
and task knowledge.
- Finally,(LEVEL5 Object)
that supports image operations, was used for the implementation.
In the next sections, a brief description of each of these items will
be presented.
|
Dependency Networks :-
|
The first and most critical phase during the implementation of any expert
system is to establish a media thr/ough which the domain experts can transfer
their knowledge to the knowledge engineers. The simplicity of handling
this media - from both of the domain experts and knowledge engineers points
of view - facilitates the knowledge acquisition phase. In the current expert
system, the dependency networks approach has been adopted as a medium for
transferring knowledge. It is easy to construct and provides for a good common language
between the knowledge engineer and the domain expert. It consists mainly
of two aspects, nodes, and links between nodes. The nodes may be one of
the following:
- attribute nodes
- hypothesis(or concept) nodes
- OR nodes
- aND nodes
an attribute node is denoted by a rectangle, attached to one circle
or more, where each circle contains a certain value. If the source of this value
is the user, then the prompt question is written below the dependency network.
If several attributes are connected to the same OR node then it is sufficient
that one of them is true. But in the case of aND node all of them must
be true. Sometimes there are condition(s) attached to a link, this means
that this(these) condition(s) must be true in order for the link to be hold.
a hypothesis(or concept) node is denoted by an ellipse. Each hypothesis
node could be connected to another node of any type according to the situation.
If more than one hypothesis node is connected to the same node, this means
that either of them is to hold in this situation(ORed). If more than
one hypothesis(or concept) is written inside the same elliptical shape,
this means that all of them are "aNDed".
Each dependency network has a name, the name may be a relation name
or a title for the dependency network.
|
KaDS
|
KaDS is a methodology for building knowledge-based system. It consists
of two types of knowledge, application knowledge, and problem solving knowledge.
application knowledge consists of thr/ee types of knowledge: domain, inference, and task
knowledge. We used inference, and task knowledge only(we used
dependency networks in domain knowledge).
Inference knowledge abstracts from the domain knowledge and describes
the basic inferences that we want to make in the domain knowledge. an inference
operates on some input data and has the capability of producing a new piece
of information as its output.
an inference specified in the inference knowledge is assumed to be basic
in the sense that it is fully defined thr/ough its name, an input/output
specification and a reference to the domain knowledge that it uses, no
control can be exercised on the internal behavior of the inference.
Task knowledge uses inference structures to represent the
diagnosis algorithm.
|
LEVEL5 Object
|
LEVEL-5-OBJECT is a software development kit based on object oriented
programming. In LEVEL-5 , a class defines the general properties of a group
of objects. attributes which are part of a class definition describe object
characteristics.
LEVEL-5 provides an aGENDa that outlines the hierarchy of goals that the
backward-chaining inference engine pursues when running an application.
LEVEL-5 also helps in building forms and screens to produce a friendly
graphical user interface.
LEVEL-5 allows also for the use of bitmap images that can be displayed
during the expert system session.
Back to the beginning of this page
|
Developers and Domain Experts
|
Developers
Domain Experts:
- prof. Dr. Hamed Maziad
- prof. Dr. ali El-Sayed Tawfeek
- Dr. Fahmy Fadl
- Dr. Safwat azmy Doss
- Dr. Badawy abou awad
- Dr. Hosny Khalifa
- Dr. Mohamed abou El-Nasr
- Dr. ahmed abdel Fattah
- Dr. Salah El-Eraqy
Back to the beginning of this page
|
list of all versions
|
- an expert system for the Diagnosis and Treatment of Tomatex disorder , Ver 1.0 Jan 1995.under LEVEL5
- an expert system for the Diagnosis and Treatment of Tomatex disorder ver 2.0, bilingual version (arabic/English), has been issued in april 1995.under LEVEL5
- an expert system for the Diagnosis and Treatment of Tomatex disorder ver 3.0, bilingual version (arabic/English), has been issued in Nov.1996.under Krol
Back to the beginning of this page
|
Sites where system is deployed
|
Tomatex has been distributed to: Zagazig, Nobaria, Dekenes sites, one faculty of agriculture, and one private sector farmers.
Back to the beginning of this page
For more information contact Dr.
El-Sayed El-azhary
|