Sie sind auf Seite 1von 41

A Rule-Based Expert System Using an Interactive Questionand-Answer Sequence

Jinmu Choi (jmchoi@arches.uga.edu) Dept. of Geography University of Geogia

Abstract
Spatial analysis in GIS commonly has been performed with statistics packages as analysis tools. Other approaches have included the use of fuzzy logic, cellular automata, neural-networks, and artificial intelligence (AI) techniques (such as knowledge-based systems). This paper examines an application of AI techniques in GIS. AI techniques can be considered as tools that emulate human thought to help in solving problems.Although some AI techniques have been applied to spatial data, it is very difficult for non-experts in AI techniques to develop and use a hybrid system based on GIS and AI techniques because of computer programming skills required to apply experts knowledge. However, a rule-based expert system uses a simple, easy rule structure, and hence has been the most common technique in knowledge-based AI for developing hybrid systems with GIS. The rules can be programmed using a simple if-then structure in a natural language such as English, Japanese, or Korean. Previous rule-based expert systems in GIS applications have used an automatic procedure for land use/cover classification. In this paper, a rule-based expert system is presented that uses an interactive question-and-answer sequence. In this system, the knowledge (prologue, questions and answers, and rules) can be constructed easily by an

expert using English. Using this system, users can determine a solution to a specific problem easily from the constructed rules through the interactive question and answer sequence. The questions and solutions will be generated by a rule-based engine using existing rules. This rule-based expert system consists largely of a main window system (expert part, user part, and map part), rulebased engine, and a database. This system is applied to the interactive interpretation of aerial photographs for neophytes.

1. Introduction
One of the most important capabilities of geographic information system (GIS) is to interpret and solve spatial problems. Generally, researchers use a GIS interface statistics packages for spatial analysis. Other approaches for the analysis of spatial data have focused on fuzzy logic, cellular automata, neural-networks, and artificial intelligence (AI) techniques (such as knowledge-based systems) with GIS. Although some AI techniques have been applied to spatial data, it is very difficult for a non-expert in AI techniques to develop and use a hybrid system between GIS and AI techniques. The reason is that most AI techniques require computer programming skills in order to apply knowledge to the system. It is not a simple process. However, a rulebased system has been the most common technique in knowledge-based AI for a hybrid system with GIS because it has a very simple rule structure to derive an answer. Once the rule-based engine (or interpreter) for the system is developed, the rule for an application can be easily constructed by a non-expert in computer programming because the rules can be programmed by a simple if-then structure using a natural language such as English, Japanese, and Korean. The purpose of this paper is to develop a simple rule-based system in which the knowledge (prologue, questions and answers, and rules) can be constructed easily by an expert using a natural language (English). Using this system, users

who are not good at both programming and problem domain can get a solution to a specific problem easily from the constructed rules through the interactive question and answer sequence. The questions will be generated by the rulebased engine using existing rules. This system is not oriented to a specific problem. Instead, it can be applied to solve various problems if the rules for those problems are already constructed in this system. In this paper, this system is applied to assist the interpretation of aerial photographs.

2.Artificial Intelligence
Artificial intelligence (AI) can be considered as tools that emulate human thought to help in solving problems. The academic definition of AI states that it is the subfield of computer science that focuses on the computers ability to manipulate nonnumeric symbols and infer new facts from sets of known facts (Carrico et al., 1989). A more practical definition is "Artificial Intelligence is the study of ideas that enable computers to be intelligent" (Winston, 1984). The goal of AI can be to make computers more useful. Carrico stated that the term artificial intelligence covers a broad range of areas (Figure 1).

Figure 1. Branches of Artificial Intelligence (AI) (Carrico et al., 1989)

The five branches of AI are derived from cognitive science (Carrico et al., 1989). Cognitive science is the field of study that seeks to understand and mimic human mental processes. This is the science that fuels the fields of AI. Robotics is the study of machines that perform mechanical manipulation and how to make them function with some "intelligence" and autonomy. Robotics is proving to be very valuable to the manufacturing industry. There are robotic automobile painters and assembly line workers. Vision systems are those that successfully interpret two- and three-dimensional pictures from a two-dimensional image obtained through manmade sensors. This involves processing the image, classifying it and then interpreting the scene. Natural language understanding is the ability to communicate with a computer by conventional language text, such as English, Russian, etc., instead of a highly structured language such as standard query language (SQL). There are a few applications for narrowly defined problems. Sound recognition is the field in which processing and reasoning about acoustic sensors such as alarms, spoken words, or automobile engines takes place. Sound recognition systems take an audible sound and make it readable. Knowledge systems are software systems that have structured knowledge about a field of expertise. They are able to solve some problems within their domain by using knowledge derived from experts in the field. This approach emphasizes data interpretation. In the five branches, experts are more interested in knowledge systems because these systems can be applied to any kinds of knowledge in order to solve problems within their domain. There are several approaches using knowledge in solving problems of various domains. These are rule-based expert systems, model-based reasoning (MBR) (Davis and Hamscher, 1986), and case-based reasoning (CBR) (Slade, 1991).

3.Rule-Based Expert System

Early programs such as DENDRAL (Buchanan et el., 1969) and MYCIN (Shortliffe, 1976) demonstrated the possibility of simulating how experts such as chemists or physicians solve their problems. After these successes, many expert systems have been developed for many technical applications of various domains.

3.1 Definition According to the initial definition of knowledge system (Carrico et al., 1989), a knowledge system is called as an expert system because this system solves problems by the knowledge of experts. The basic unit of knowledge in these expert systems is the rule, defining so-called rule-based expert system. A rule is a conditional test-action pair, for example, if condition is true, then action.

3.2 Architecture of Expert System An expert system has four main architectural components that are the knowledge base, the inference engine, the knowledge acquisition module, and the user interface for input/output (Nikolopoulos, 1997) (Figure 2). The knowledge base contains the domain specific knowledge that is used for problem solving in the domain. Knowledge can be represented and stored in the knowledge base with a format suitable for computer manipulation. One of the most commonly used ways to represent knowledge is as rules (in the form of If-then).

Figure 2. The Architecture of an Expert System The inference engine is based on an inference rule and a search strategy and contains algorithms. Algorithms are used to manipulate the knowledge stored in the knowledge base in order to solve problems. An inference rule is a way to deduce new knowledge from the existing knowledge base of rules and facts. The knowledge acquisition module enables experts to store their knowledge in the knowledge base or expert system to deduce new knowledge from existing knowledge through a machine learning process. The interface for input/output is used for the expert system to interact with the user, the environment and other systems such as databases, and spreadsheets.

3.3 Inference Method Rules have enabled experts and AI researchers to represent their problem-solving knowledge as models that could be implemented computationally (Watson, 1997). In rule-based systems, knowledge is represented as facts about the

world and rules to manipulate the facts. At any one time more than one rule may be applied to solve a problem (Figure 3) and when each rule is applied other rules may applicable to those rules (Figure 4). Therefore, a rule-based system needs a control structure to decide which rule should be applied first or next and which rules are put together.

Figure 3. Simple rule (Watson, 1997) In simple set of rules, if A is true, it can be logically inferred that D is true by using the three rules. It is enough for a rule-based system to decide the order of applied rules.

Figure 4. Complex rules (Watson, 1997) In a complex set of rules, if A is true, B and C are logically true at the same time. However, in a computer it is difficult for B and C to be true at the same time, and it has to be decided which one is true first. To solve this problem Watson discussed depth-first forward chaining and backward chaining. In fact, it is up to the programmer to decide which one is true first, but depth-first forward chaining, a commonly used control structure, would mean that inferences would be made in the order indicated by the numbers on left flow in Figure 4 (Watson, 1997). Depth-first forward chaining uses the depth of the tree structure of condition-action relation. If F was true and a solution before it is inferred that E is true, it may not need to infer that E is true. This forward chaining is used to infer new facts from existing facts. It is also possible to use the same set of rules in reverse to determine what needs to be true for a premise to be true. This method is called backward chaining. Backward chaining is commonly used in rule-based expert systems to enable a hypothesis to be tested and this problem-solving strategy is often referred to as generate and test (Watson, 1997). When

your car is out of order and you go to mechanic, if it is a complex problem the mechanic will try to confirm a hypothesis by looking for characteristic symptoms or by performing certain tests. If these tests do not confirm the hypothesis, the mechanic will think of another problem and perform tests for that hypothesis. This is backward chaining.

3.4 Advantages and Limitations of Rules Rule-based systems which have several advantages have been developed successfully (Chadwick and Hannah, 1986, and Watson, 1997). First, rules can be easily understood by programmers or experts because the control structure is relatively simple as an if-then format. Second, rules can encapsulate an important part of knowledge in that rules can model a complex problem. Also, knowledge can be saved by rules. Third, the control structure of a rule seems to mimic some human problem-solving strategies, and hence rules saved in an expert system can be used for solving human problems. Fourth, rules are independent of each other. Even though some parts of the rules are deleted or incorrect, these mistakes will not affect all rules or functions. Some mistakes of other programming languages that can be used to describe an if-then structure may affect the entire function. Fifth, rules can be placed in any order in a program. Last, natural languages can be used for questions on the user interface, and hence it is very convenient for non-experts to use rule-based expert system for solving problems. However, there are some limitations to rule-based systems (Slade, 1991, and Watson, 1997). Sometime, it is difficult to obtain a correct set of rules. There are several causes. First, it may take long time for an expert to give knowledge to a programmer if the rule cannot be constructed by a natural language. Second, even though an expert gives knowledge to a programmer, it may be difficult for the programmer to understand the knowledge fully and produce a correct model. Third, if typical diagnostics or complex tasks should be executed, several hundred rules may be required. Fourth, although a rule assumes that there is a generally accepted body of explicit knowledge in a domain, there are many domains in a real world that have no underlying causal models and no generally accepted explicit rules.

3.5 Software for Expert System Development There are four kinds of software tools available for rule-based expert system development: conventional languages (such as Basic, Pascal, and C), AI languages (such as LISP and Prolog), AI development environments (such as automated reasoning tool (ART) and knowledge engineering environment (KEE) (Chris, 1997)), or AI shells (such as Savoir) (Bramer, 1989). AI languages have the most flexibility for development of an expert system but require extension for implementing facilities such as the user interface and debugging tools. AI development environments have many facilities for creating the user interface and tools for debugging. However, AI development environments are still very large and expensive. Carrico (1989) recommended using rule-based shells for knowledge systems development. Shells are a form of reusable code: not binary but script. Shells provide preconstructed, essential utilities such as user interfaces, report generators, and the part of the program that scans, selects, and applies rules and data. Rule-based shells are a kind of database generator. Therefore, the shell handles the data dictionary, operating stack, concurrent calculations, I/O functions, audit and recovery functions, and etc. automatically and in the background. However, Bramer (1989) argued that developing a fully operational system from a prototype often requires complete recoding using an entirely different software package although AI shells are fast and convenient for developing simple and small expert systems such as prototypes. Conventional languages have been rarely used for developing expert systems because it is difficult to develop the complex reasoning process of a huge expert system. However, some conventional languages such as visual C++, Java, or visual Basic have well designed facilities for developing a user interface. If an expert system has to cooperate with other non-expert systems, then the expert system should be extendable using one of conventional languages because most non-expert systems are developed using a conventional language. Therefore, conventional languages may be advantageous over the other three expert system development tools when a certain expert system has a simple straightforward reasoning engine and needs to cooperate with other non-expert system.

4. System Development

This rule-based system consists largely of a main window system (knowledge acquisition module, rule-based inference engine, and user interface for input/output), a simple GIS mapping system, and a database. The knowledge acquisition module, inference engine, and user interface were built using Visual Basic 6.0 (Harris, 1997, Ryu, 1999, and Stephoens, 2000). The GIS mapping system was also built using Visual Basic 6.0 based on MapObjects Version 2 (ESRI, 1999). MapObjects has many GIS facilities and can be extended and integrated easily with other systems using a conventional language such as Visual C++ and Visual Basic. The database for the knowledge base (prologue, questions and answers, and rules) is Microsoft Access 2002.

4.1 System Design The main window system consists of an expert part, a user part, and a map part (Figure 5). The main menu of this system consists of these three parts. The expert part is the knowledge acquisition module. The expert part can be used to construct new knowledge (prologue, question, answer, and rule) or to update the existing knowledge by experts. The user part is the user interface for the input/output of the knowledge base. The user part can be used to solve a specific problem from the existing knowledge in the database using an interactive question and answer sequence. The map part is a simple GIS mapping system that is integrated into this rule based expert system. The map part can be used to analyze any kind of vector and raster (image) file for visual interpretation, or to draw a thematic map using drawing tools.

Figure 5. Main window of rule-based system

4.1.1 Expert Part The expert main menu consists of three sub menus: new problem, open problem, and exit (Figure 6). The new problem (Figure 7) menu can be used to construct new knowledge (prologue, question and answer, and rules).

Figure 6. Submenu of expert system

Figure 7. New problem window of the expert menu

Figure 8. Input window for new question in the new problem window The new problem window has several components: text boxes and buttons. The problem text box is used to enter the name of the problem using a natural language (English). The other text boxes are used to enter the contents of the knowledge. The buttons have own functions.

Table 1. Function of each button of new problem window


Button Data input Save Print Exit Function To open the input window (Figure 8) for knowledge To save the knowledge putted To print the content of the knowledge To close new problem window

The open problem window (Figure 9) can be used to revise or update the existing knowledge (prologue, question, answer, and rules). It also has several components: a combo box, text boxes and buttons.

Figure 9. Open problem window of the expert menu The problem combo box is used to select the existing name of problem. The text boxes are used to show the existing contents of the knowledge. The buttons have their own functions.

Table 2. Function of each button of open problem window


Button Data update Print Exit Function To open the update window (Figure 10) for knowledge To print the content of the knowledge To close new problem window

Figure10. Update window for the existing questions in the open problem window

4.1.2 User Part

The user main menu has one sub menu: select problem (Figure 11). The select problem menu will open the solving problem window (Figure 12). This window has a combo box, text boxes, and buttons.

Figure 11. Submenu of user part

Figure 12. Solving problem window for the user menu The select combo box can be used to select the existing problem. The text box for prologue is used to show the prologue (or explanation) for the selected problem. The text boxes in the question and answer tab can be used to solve the problem by a user interactive process. The rule-based engine will generate the question in the question text box. For

this question, the user has to answer that question using answer text box. Once the problem is solved through the user interactive process, then the result will be shown in the solution text box. The print and close buttons can be used to print the contents and to close the window, respectively. The rule-based inference engine was implemented in the user part because the inference rule and the search strategy are needed in the user part in order to solve a selected problem. There are three types of inference rules: deduction, induction, and mixture of deduction and induction. There are also three types of search strategy: forward chaining, backward chaining, and mixture of forward and backward chaining. This inference engine is built based on deductive reasoning using forward chaining. The inference engine will generate questions automatically when the user selects a problem. The solution will also be generated from the previous questions and answers using the existing rules by the rule-based inference engine.

4.1.3 Map Part The Map main menu has one sub menu: map viewer (Figure 13). The map viewer menu will open the map viewer window for the visual interpretation of GIS data (vector or raster image) and mapping for the thematic map.

Figure 13. Submenu of map part

This map viewer can support raster data (ESRI Grid data), standard image data (bitmap (*.bmp), gif (*.gif), jpeg (*.jpg), window metafile (*.wmf), and so on), and several vector data formats (ESRI coverage, ESRI shape data, and CAD drawings). The open layer module in the map viewer is used for loading several vector or raster layers in the map canvas (Figure 14).

Figure 14. Map Viewer window for the visual interpretation or mapping

The map viewer can be used for visual interpretation of GIS data using scaling modules (Zoom In/Out and Pan). Several GIS data layers can be overlaid for the same area. For vector data, the symbol property window can help the user change and classify data for improving interpretation. The identify module (the i button on the menu bar) can help the user identify feature properties in vector data. The map viewer also can be used for the thematic mapping (Figure 15). The thematic mapping process has four steps. First, the user selects features on the existing image or map. Second, the user draws polygons for the boundary of the selected features. Third, the user interprets the selected features using the user part in the expert system. Last, the user labels the polygon for selected features using labeling module when the expert system generates the solution.

Figure 15. Map Viewer window for thematic mapping

4.2 Database Design The relational database structure (Figure 16) is used to save the knowledge for problems. The database includes a problem table, a question table, and a rule table. The problem table contains the name and prologue for all problems. The question table contains both questions and answers for one problem. The question is symbolized using simply one word for easy processing in the rule-based inference engine. The rule table contains the rules for one problem. In Figure 16, each problem in the problem table is related to the question table and the rule table for that problem. If a new problem is created, the problem will be added in a new row in the problem table. And, the question and rule tables for that problem will be created in the database.

Figure 16. Table structure and relations in the database

5. Interpretation of Aerial Photographs


The objective of this application is to construct rules for the interpretation of aerial photographs. For this purpose, the elements for land use/cover interpretation of a black and white aerial photograph and how to use these elements for constructing rules and symbols for questions and answers in this rule-based system should be understood. In this application, eleven land use/cover categories were included to develop a rule: residential, commercial and services, industrial, highway, railroad, deciduous forest, evergreen forest, mixed forest, lakes, school, and golf course. The site of this study is the Navy Supply School area of Athens, GA, and the scale of the aerial photograph is 1:18,000. There are seven basic elements for the interpretation of aerial photographs: tone, texture, pattern, shape, shadow, size, and situation (Avery and Berlin, 1992). Tone is the density of brightness. It is a record of light reflection from the land surface on the photo. Texture is a frequency of tone change within the photographs. Pattern is the spatial arrangements of objects. Shape is the general form of an object. Shadow reveals shape and height of an object. Size of an object is related to that of the others. Situation is the position of an object in relation to those in the immediate vicinity.

5.1 Decision Tree and Rule Among the seven elements of interpretation of an aerial photograph, six elements are used in this application: tone, texture, pattern, size, shape, and situation. Shadow cannot be used in the interpretation because of the small scale of the aerial photograph. Using these six elements, rules in this rule-based system were constructed. First, situation is used to separate urban uses from others such as lakes and forests. If there are artificial structures such as buildings or houses, this area may be used for urban land use/covers such as commercial, industrial, schools, residential, and transportations. In the urban uses, some categories are separated from each other using shape and situation. The forest area is mainly separated by the tone.

Figure 17. Decision tree for interpretation of aerial photographs The most difficult categories for interpretation are highways and railroads because of the small scale of the aerial photograph. However, highways are wider than railroads or small roads. Also, railroads can be separated from small roads between residential areas because railroads have few intersections but roads have many intersections. The decision tree of the rule is shown in Figure 17. From this decision tree, the knowledge (prologue (appendix I), questions and answers (appendix II), and rules (appendix III)) for interpretation were constructed in the database.

6. Discussion
There are some limitations in applying this rule-based system. First, if a typical diagnostic or complex task has to be executed, several hundred rules may be required. If the rules need much knowledge in various domains, it may be very difficult to construct the rules because of different aspects among domains. Second, a rule-based expert system cannot use experiences that have been used to solve the similar problem. Even though a problem is the same as or similar to the old one, users have to answer again all questions in this rule-based system. There is no way to save a similar case in the rule-based system. The maintenance of knowledge in a rule-based system is normally a manual process requiring further knowledge acquisition. Third, there are no exceptions in rules. Rules need explicit knowledge, exact questions, and exact answers. It is not allowed for users to make mistakes in answers. To overcome these limitations, there are some other approaches in artificial intelligence: model-based reasoning (MBR) (Davis and Hamscher, 1986) or case-based reasoning (CBR) (Slade, 1991). However, MBR also has some limitations (Davis and Hamscher, 1986). First, the current modeling technology cannot model and predict the subtle and complicated interactions in certain domains. Second, simple problems dont have to be modeled using complex modeling technology. Therefore, MBR should be used under two assumptions. First, the

structure and behavior of the device should be reasonably well known and simple enough to model, but complex enough that exhaustive simulation is impossible. Second, the set of possible faults should be difficult to reliably enumerate in advance. Therefore, CBR is commonly used in recent research of AI. There are several advantages of CBR over rule-based systems (Holt et el, 1997). First, CBR can handle domains where problems have many exceptions to rules and where problems are not fully understood. Second, cases learn from experiences, that is keep up with knowledge that workers learn in their daily experiences, indicating an ability to store temporal information. Third, a case represents the experts knowledge more accurately (Schank, (1982) suggested that experts usually tend to use knowledge in the form of particular experiences (cases) rather than in the form of rules). Fourth, CBR provides methodologies for validation and maintaining the knowledge (case) and application. Last, CBR offers cost-effective solutions to knowledge acquisition bottleneck problems (Slade, 1991). However, rule-based systems are more suited to solve a problem when it is difficult to gather case data (Althoff et al. 1994). In an expert aspect, the knowledge can be more easily implemented in rule-based systems than in CBR systems because rules can be constructed by very simple reasoning structure (if-then structure) and a natural language.

7. Conclusion
In this paper, the rule-based expert system was developed. This system consists of three main parts: expert, user, and map. The expert part can be used to construct new knowledge (prologue, questions and answers, and rules) or update existing knowledge. The user part can be used to solve a certain problem for which the rules are already constructed in this system. The map part can be used to load GIS data layers. Various vector and raster file types can be loaded. Especially, map part will be helpful for visual interpretation of an image data. In remote sensing and GIS, previous studies (Tatsuyama, 1987, Moller-Jensen, 1990, Foschi and Smith, 1997, Wilson, 1997, and Usery et al., 1988) using rule-based expert systems focused on the automatic interpretation of aerial

photographs or satellite images or on the automatic mapping of spatial data for experts. However, in this study, the rule-based expert system was applied to the interactive interpretation of aerial photographs for neophytes. The interpretation of aerial photograph is not an easy task for the neophytes in aerial photography. If users cannot identify a certain feature during the interpretation of aerial photographs, users can consult it from this rule-based system through the question and answer sequence. The question will be generated automatically by the users answer. The knowledge was constructed for the aerial photograph of the study area using the interpretation elements in aerial photography. Using this rule-based system, eleven land use/cover categories were interpreted easily from the aerial photograph. For further study, the map part in this system can be enhanced into mapping or digitizing system of GIS. Then, users can be consulted by this expert system during mapping or digitizing in order to produce GIS data or map products.

Appendicies

Appendix 1. Prologue for Aerial Photograph Interpretation THIS RULE IS DESIGNED FOR INTERPRETATION OF A BLACK AND WHITE PHOTO. WITH THIS RULE, YOU CAN INTERPRET URBAN LAND COVER AND LAND USE (J.R. ANDERSON ET AL.(1076)'S LAND USE AND LAND COVER LASSIFICATION SYSTEM LEVEL II).

FOR GETTING AN ANSWER, YOU MUST ANSWER ALL QUESTIONS USING FOLLOW INTERPRETATION ELEMENTS.

1. TONE IS A RECORD OF LIGHT REFLECTION FROM THE LAND SURFACE ON THE FILM: DARK OR LIGHT 2. TEXTURE IS A FREQUENCY OF TONE CHANGE WITHIN THE IMAGE: COARSE OR FINE 3. PATTERN IS SPATIAL ARRANGEMENTS OF OBJECTS: REGULAR OR IRREGULAR 4. SHAPE IS A GENERAL FORM OF AN OBJECT: CIRCLE OR RECTANGULAR OR OTHER 5. SHADOW REVEALS SHAPE AND HEIGHT OF AN OBJECT 6. SIZE OF OBJECT IS RELATED TO THAT OF THE OTHER: LARGER OR SMALLER 7. SITUATION IS THE POSITION OF AN OBJECT IN RELATION TO THOSE IN THE IMMEDIATE VICINITY

Appendix 2. Questions and Answers for Aerial Photograph Interpretation Symbol: STRUCTURE Question: Is there artificial structure? YES OR NO Symbol: SIZE Question: Is the structure small? YES OR NO

Symbol: SHAPE1 Question: Is the structure circle? YES OR NO Symbol: SITUATION1 Question: Is there playground adjacent to the structure? YES OR NO Symbol: SITUATION2 Question: Is there large parking lots adjacent to the structure? YES OR NO Symbol: SITUATION3 Question: Is there some large structures near to the structure? YES OR NO Symbol: SITUATION4 Question: Does the structure have flat roof? YES OR NO Symbol: SHAPE2 Question: Does it look like long linear feature? YES OR NO

Symbol: SHAPE3 Question: Is the line feature wider than the other line features? YES OR NO Symbol: SITUATION5 Question: Is there many intersections / structures along the line feature? YES OR NO Symbol: PATTERN Question: Is it stripped irregularly with dark and white tone? YES OR NO Symbol: SHAPE4 Question: Is the surface of the feature smooth? YES OR NO Symbol: TONE Question: Is the tone darker than other similar features (WHOL OR PART OR NO)?

Appendix 3. Rules for Photograph Interpretation

If
STRUCTURE EQ YES AND SHAPE2 EQ NO AND SIZE EQ YES AND SHAPE1 EQ NO AND SITUATION2 EQ NO STRUCTURE EQ YES AND SHAPE2 EQ NO AND SIZE EQ YES AND SHAPE1 EQ NO AND SITUATION2 EQ YES STRUCTURE EQ YES AND SHAPE2 EQ NO AND SIZE EQ YES AND SHAPE1 EQ YES STRUCTURE EQ YES AND SHAPE2 EQ NO AND SIZE EQ NO AND SITUATION1 EQ YES AND SITUATION2 EQ NO STRUCTURE EQ YES AND SHAPE2 EQ NO AND SIZE EQ NO AND SITUATION1 EQ YES AND SITUATION2 EQ YES STRUCTURE EQ YES AND SHAPE2 EQ NO AND SIZE EQ NO AND SITUATION1 EQ NO AND SITUATION2 EQ NO STRUCTURE EQ YES AND SHAPE2 EQ NO AND SIZE EQ NO AND SITUATION1 EQ NO AND SITUATION2 EQ YES AND SITUATION3 EQ NO AND PATTERN EQ NO STRUCTURE EQ YES AND SHAPE2 EQ NO AND SIZE EQ NO AND SITUATION1 EQ NO AND SITUATION2 EQ YES AND SITUATION3 EQ NO AND PATTERN EQ YES STRUCTURE EQ YES AND SHAPE2 EQ NO AND SIZE EQ NO AND SITUATION1 EQ NO AND SITUATION2 EQ YES AND SITUATION3 EQ YES AND SITUATION4 EQ NO STRUCTURE EQ YES AND SHAPE2 EQ NO AND SIZE EQ NO AND SITUATION1 EQ NO AND SITUATION2 EQ YES AND SITUATION3 EQ YES AND SITUATION4 EQ YES STRUCTURE EQ YES AND SHAPE2 EQ YES AND SHAPE3 EQ YES STRUCTURE EQ YES AND SHAPE2 EQ YES AND SHAPE3 EQ NO AND SITUATION5 EQ NO STRUCTURE EQ YES AND SHAPE2 EQ YES AND SHAPE3 EQ NO AND SITUATION5 EQ YES STRUCTURE EQ NO AND PATTERN EQ YES

Then
Residential Commercial Industrial School Service (park) Commercial Commercial Golf Course Residential (Apartment) Industrial Highway Railroad Small Car Road Golf Course

If
STRUCTURE EQ NO AND PATTERN EQ NO AND SHAPE4 EQ YES STRUCTURE EQ NO AND PATTERN EQ NO AND SHAPE4 EQ NO AND TONE EQ WHOLEY

Then
Lake Evergreen

Acknowledgments
I wish to thank Dr. Lynn Usery who reviewed and commented on this paper.

Reference
Althoff, K., S. Wess, and M. Manago, 1994, "Reasoning with Cases-Theory and Practice", Proceedings of the European Conference on Artificial Intelligence, pp. 1-26. Avery, T.E. and G.L. Berlin, 1992, Fundamentals of Remote Sensing and Airphoto Interpretation, 5th Ed, Macmillan Publishing Company, New York, pp. 52-57.

Bramer, M., 1989, "Ch3. Expert systems: where are we and where are we going?", In Forsyth, R., (eds), 1989, EXPERT SYSTEMS: Principles and case studies, 2nd Ed, Chapman and Hall Ltd, pp. 31-53. Buchanan, B., G. Sutherland, and E. Feigenbaum, 1969, HEURISTIC DENDRAL: A program for Generating Explanatory Hypotheses in Organic Chemistry. Machine intelligence 4., Edinburgh University Press. Carrico, M. A., J. C. Girard, and J. P. Jones, 1989, Building knowledge systems, McGraw-Hill Book Co., New York. Chadwick, M. and J.A. Hannah, 1986, Expert Systems for Personal Computers - An Introduction to Artificial Intelligence, Sigma Press, Wilmslow, England, Chapter 2: Principles of Rule-based Expert Systems, pp.9-29. Chris, N, 1997, "Appendix C: PROLOG and Expert System Products", Expert Systems: Introduction to First and Second Generation and Hybrid Knowledge Based Systems, Marcel Dekker, Inc., New York, pp.317-326. Davis, R., and W. Hamscher, 1986, "Model-based Reasoning: Troubleshooting", Exploring Artificial Intelligence, Morgan Kaufmann publishers, Inc., pp. 239-346. Environmental Systems Research Institute, 1999, ESRI MapObjectsTM Version 2.0, Environmental Systems Research Institute, Inc. Foschi, P.G. and D.K. Smith, 1997, "Detecting Subpixel Woody Vegetation in Digital Imagery Using Two Artificial Intelligence Approaches", Photogrammetric Engineering and Remote Sensing, Vol. 63, No. 5, pp.493-500. Holt, A. and G. L. Benwell, 1997, "Applying case-based reasoning techniques in GIS", International Journal of Geographical Information Science, vol. 13, no. 1, pp. 9-25. Harris, M., 1997, Teach yourself: Visual Basic in 21 Days for Applications 5, SAMS. Moller-Jensen, L., 1990, "Knowledge-Based Classification of an Urban Area Using Texture and Context Information in Landsat-TM Imagery", Photogrammetric Engineering and Remote Sensing, Vol. 56, No. 6, pp.899-904.

Nikolopoulos, C., 1997, Expert Systems: Introduction to First and Second Generation and Hybrid Knowledge Based Systems, MARCEL DEKKER, INC., New York. Ryu, J., 2000, Korean Visual Basic 6.0 using Bible, Youngin.com, Seoul Schank, R., 1982, Dynamic Memory: A Theory of Learning in Computers and People, Cambridge University Press. Shortliffe, E., 1976, Computer-Based Medical Consultations: MYCIN, American Elsevier Co., New York. Slade, S., 1991, "Case-Based Reasoning: A Research Paradigm", AI MAGAZINE, vol. 12, pp. 42-55. Stephens, R., 2000, Visual Basic Graphics Programming, 2nd Ed, John Wiley & Sons, Inc., New York. Tatsuyama, T., 1987, "Knowledge-Based Aerial Image Understanding Systems and Expert Systems for Image Processing", IEEE Transactions on Geoscience and Remote Sensing, Vol, GE-25, No.3, pp305-316. Usery, E.L., P. Altheide, R.R.P. Deister, and D.J. Barr, 1988, "Knowledge-Based GIS Techniques Applied to Geological Engineering", Photogrammetric Engineering and Remote Sensing, Vol. 54, No. 11, pp.1623-1628. Watson, L., 1997, Applying Case-Based Reasoning, Morgan Kaufmann Publishers, Inc. Wilson, P.A., 1997, "Rule-Based Classification of Water in Landsat MSS Images Using the Variance Filter", Photogrammetric Engineering and Remote Sensing, Vol. 63, No. 5, pp.485-491. Winston, P.H., 1984, Artificial Intelligence, 2nd Ed., Addison-Wesley publishing Co.

-END-

ACTION PLAN

Das könnte Ihnen auch gefallen