GISdevelopment.net --> Application --> Urban Planning
IRS-1B based FCC image interpretation for land use and land cover mapping: An expert system approach
Rajkishore Prasad
Lecturer, University Deptt. of Electronics, B. R. A.
Bihar University, Muzzafarpur
rkishore2k@hotmail.com
A. K. Sinha
Professor Department of Electrical Engineering,
MIT, Muzzafarpur
aksinha_1@yahoo.com
Rajkishore Prasad
Lecturer, University Deptt. of Electronics, B. R. A.
Bihar University, Muzzafarpur
rkishore2k@hotmail.com
A. K. Sinha
Professor Department of Electrical Engineering,
MIT, Muzzafarpur
aksinha_1@yahoo.com
Introduction
The science and art of remote sensing has emerged as one of the most important application of space technology in obtaining environmental and natural resources data related to earth using various aerial platforms. These data are being used in large number of areas e.g. agriculture, forestry, archeology, geography, geology, oceanography, ecology etc to obtain useful information for the future planning, management and sustainable development of the natural resources. Such data are collected using different aerial platforms. However, the satellite based remote sensing has become very popular and different countries have launched remote-sensing satellites, for this purpose. The collected data are processed and interpreted in different forms and formats using digital techniques or optical techniques to extract useful information. One of the most widely used data format for information extraction is the infrared False Color Composite (FCC) image. The extraction of information from such images about ground reality is done by image interpretation for which generally three methods namely photo interpretation, spectral analysis and data integration are used. However, for the public, satellite images and the information they bear seem vague and far beyond the understanding of a layman. In the coming decades remote sensing will be one of the key tool for making critical decisions affecting the earth, its environment and resources. So it is important that the informed citizens should have a basic understanding of these technologies and how they are used. Such awareness in mass will ensure the credibility and faith in the accuracy of information derived from the remote sensor data.
In this paper our main concern is with photo interpretation method of IRS-1B standard FCC and to describe the development of a rule based expert system, using expert system shell ESTA (Expert System shell for Text Animation),(Fig.1) for the beginners to learn the visual interpretation skill for the purpose of land use & land cover mapping.
Fig. 1: Expert system shell to expert system
Expert system and FCC image interpretation
Photo interpretation is the visual interpretation of images based on features tone, pattern, shape, size, shadow, texture and association. Most of the conventional digital image processing techniques is based on color or size or texture or tonal variation of each pixel in the image. In contrast to digital analysis of the images, a human interpreter does not interpret the image pixel by pixel. Instead he or she exploits the aggregate information related to various basic image-features of unknown objects along with his scientific knowledge, general knowledge of the phenomena as well as experience to do classification [1,2]. As a consequence, the interpretation result for land use and land cover produced by a well-trained human interpreter is often less crude than the same obtained using digital techniques [3]. For human interpreter it is easy to interpret natural color image but the interpretation of FCC image becomes difficult and requires adequate training and experience [4]. Also different band combinations of satellite data for three primary colors result in different FCC images, which are suitable for different application. Every application of remote sensing deals with a specific subject or integrated process of different subjects. Thus the process of visual interpretation of wide variety of remotely sensed data is a complex intuitive process of combining evidential information from different sources and subjecting such information to an expert’s knowledge, experience and heuristics at each levels namely detection, identification, analysis, recognition and classification of the process. It calls for analysis of a number of related information by a domain expert. So even with on going advances in digital image processing technique the importance and role of human photo interpreter can’t be denied and it is required to train human resources in this skill. Unfortunately, this art has not been taught systematically in the related educational institutions from the last 20 years. As a result this art is being lost [5]. However, in future the scope of on-screen interpretation of high resolution remote sensing data will increase and image identification system providing way to combine together human interpreter and machine interpretation will be required as the maximum interpretation accuracy achieved using digital image processing technique has been reported up to 70 to 80% [6]. The associated information and the logical reasoning that are used by a well-trained human interpreter can be encoded in the from of rules and facts to create knowledge based systems. The need of expert level performance in the remote sensing image interpretation has brought in a shift from domain independent statistical methods to domain specific knowledge based techniques [7]. Thus the activity of image interpretation has similarity with the nature of explorative and qualitative reasoning in the line of artificial intelligence and expert system. Since the interpretation of each FCC images requires different skill and experience so for a human interpreter it becomes difficult to manage properly application of huge knowledge while making decision in photo interpretation. Also the human interpreter may be absent or not easily available. The interpretation by a human is also effected by his mental status like biases, anxieties, fatigue, repeatability of task etc. So the need of computer based support is obvious. The knowledge used in photo interpretation can be represented in logical paradigm that makes prolog [8] based expert system shell ESTA suitable for expert system development. Lot of image interpretation system has been developed in the past. These system are different in number of ways like the knowledge representation scheme, control strategy, application area etc. Most of them are based on spectral features of the image. Very few systems have been developed for developing visual interpretation skill. In the field of land use /land categorization using satellite data a rule based expert system [9] was developed using visual interpretation keys developed by NRSA (National Remote Sensing Agency), India. In other development an expert system for “Land Form Study” has been created, in the Prolog language, for landform type identification and classification in automated and user interactive way [10]. Majority of systems have been developed for automatic image interpretation e.g. the system SPAM [11] for aerial image interpretation is based on production rules. The system MESSIE [12] has used production rules as well as semantic nets. In AIDA system [13] semantic nets and rules has been used to represent knowledge base while ERNEST [14] uses semantic net only. SIGMA [15] contains three expert modules based on frame and rule representation. Recently Jenson et al (16) in his neural network based photo interpretation approach have emphasized use of hybrid approach i.e. visual interpretation and digital interpretation. However, in the neural network based system the classification rules are hidden in its weight and is not useful for a user for learning point of view.
This paper is organized as follows. In the following section and its subjection the proposed expert system development process using ESTA has been described. In the section 4, consultation with the system and result has been presented using standard FCC images produced by IRS-1B & 1D satellites data. The last section contains conclusion and future work.
Fig. 2: Section tree of the system
Expert System Development in ESTA
ESTA
Expert system came out as the first commercial fruit of AI. These are domain dependent computer programme that use knowledge and inference procedure of domain specific human expert to mimic his/her problem solving capability and way in which they solve [17]. In the beginning the expert system development was mainly carried out by the AI people using sophisticated inference engine and search techniques. However, later it was found that the expert system developed by non-AI (domain expert) people using simple inference engine and search techniques outperformed those developed by AI people. This led the AI people to realize the fact that the quality of knowledge highly influences the quality of expert system, which in turn led to the separation of knowledge module from inference and control strategies [18]. This gave a great leap in the expert system technology and resulted in the development of expert system shell. An expert system in the context of expert system shell may be defined as follows:
EXPERT SYSTEM=KNOWLEDGE BASE + EXPERT SYSTEM SHELL (Control &Inference +User Interface).
The separation of the control and inference parts in the knowledge base is a fundamental feature of an expert system shell. ESTA is also an expert system shell developed by PDC (Prolog Development Center), Denmark [19]. By providing with it a knowledge base ,as shown in fig.1, of a specific domain an expert system can be created. Thus
KNOWLEDGE BASE +ESTA= EXPERT SYSTEM.
Fig. 3: A Screen shot when user seek help during consiltator
ESTA system has several advantages over other expert system shell like CLISP. According to PDC, ESTA is easy to use and a great stand-alone environment for constructing advisory and decision support systems. Building advanced knowledge base with ESTA requires no previous programming experience and it is suitable for many problem domains. ESTA is the perfect tool for structuring of knowledge. It includes explanation facilities for the question being asked and for given advice. Details about ESTA can be found in the reference [19] or PDC homepage. It is very easy to use, fast and requires no programming experience. The knowledge base of the ESTA consists of rules represented in its own syntax. It has all the inbuilt facilities to write the rules that build the knowledge base.
Table 1
Knowledge Base Development
Palpably as an expert system shell, ESTA for knowledge engineer works as tool to create knowledge base and from user side it is used to consult the created knowledge base as shown in the fig.1.Thus the task of developing expert system using ESTA, domain independent module, is centered at developing a domain specific knowledge base, which is a multi-step process. These steps are problem identification, knowledge acquisition and knowledge representation in accordance with the syntax of ESTA. Problem identification is already contained in the previous sections. The rest two are described below.
Knowledge Acquisition
The knowledge required for present study has been gleaned out from available standard literature [3,4,20] and human experts from remote sensing center, Patna using IRS-1B based FCC images. The theme of such interpretation rules, in very brief, for land cover and land use identification is presented in the table 1
The IRS-1B satellite uses Linear Imaging Self Scanner (LISS) I & II sensors, which are approximately equivalent to Thematic Mappers (TM) bands 1,2,3& 4. . So the same knowledge base can also be used to interpret TM based standard FCC image with little variation
Knowledge Representation
The knowledge base for ESTA consists of simple if- then rules. It is also obvious from above table of visual image interpretation key that this domain is based on the specific rules for specific land use/land cover category, so the rule based approach [21] for knowledge representation will be best suited. There are two major knowledge representations in ESTA namely Section and Parameters. Section is top level of knowledge representation and contains the logical rules that directs ESTA how to solve problem, actions to perform such as giving advice, going to other section, calling to routines etc. The first section is always named as start section. Parameters are used as variable and it decides the flow of control among the sections. ESTA accepts four types of parameters namely Boolean or Logical, Text, Number and Category parameters. These parameters serve different purposes. The Boolean parameter is used when the answer to asked question is either Yes, No, or Unknown. Text parameters are used for text object. Number parameter is used to represent numerical values. Category parameters are used when variable takes more than one values. The value for any of the parameter is calculated from end-user’s response to a question, through other parameters or as a result of application of rules. Any parameter consists of declaration field, type field and number of optional field such as explanation field, rules field, picture field, question field etc.
Accordingly, the knowledge base has been developed here too. Three layers of sections for each land use and land category occurring in table1 have been used. These are shown in the fig2, which represents screen shot of the section tree. The rules in each section are arranged in the first come first serve sequence that decides flow of control among the sections. The firing of a rule in any section is governed by the related parameters. All types of parameters have been used in developing knowledge base as per the design requisition. The section consult again represents end of session and can direct control, on user’s response, such that it become possible to do consultation for all land use & land cover types in sequence in each session, otherwise, it needs to start re-consultation for each type. In order to tackle the uncertainty related with any Boolean parameter (when a user responses as Unknown to a question ) multilevel dialog has been designed so that user become aware with what is being asked by the system. The top level of dialog uses standard scientific terms and on each lower level effort has been made to simplify the terms. The declaration field of each parameter contains proper explanation for HOW type explanation while the explanation field contains the explanation to the question related to parameters. Each parameter contains picture name in the picture field, which displayed during consultation. Pictures are stored in picture database in BMP format. On the picture displayed with question of parameter, the land use land cover type has been specified using hotspot editor. Current version of ESTA allows use of picture in the title, parameter, in advice and in action list. However it has been found essential in, some places, to include pictures in explanation. This has been achieved by directing control flow in the advice and pictures have been used to make easier understanding the question. Such an example is shown in fig 3 where the user selects help option. This drives the system in advice window, as shown in the screen shot of fig.4, where the user gets option to see more interpreted FCC images to develop idea about visual image interpretation. The developed knowledge base has been stored in compiled format to optimize the performance
Fig. 4: A Screen shot in response to fig.3
Consultation and Result
The consultation with the system can be initiated by loading knowledge base in memory. ESTA allows only one knowledge base at a time. As the programme runs ESTA goes in start section and looks for the rules. It asks question related to parameters in ESTA Consult window It executes rules in order of the arrangement. The parameter Land Cover places menu of all the land cover and land use types from where user can choose any one and accordingly, the control is transferred to the related section at the second level. At the second level image interpretation is done. User gets here decision of system in advice window. However, as shown in screen shot of fig. 4, the advice window has been also used to provide help on question using additional figures for aforesaid reasons. The interpretation rules in the sections, at second level, have been placed in first come first serve sequence. These sequences have been logically arranged keeping in view minimizing the situation of conflict, which is very common in image interpretation [22]. However, this does not provide strong strategy to overcome the problem of conflict resolution.
Fig. 4: Standard FCC image taken by IRS - 13
In the present work two FCC images, taken by IRS- 1D & IRS-1B satellites, of Hajipur area of district of Vaishali of Bihar have been used for the purpose of identification of the land use/land cover. The photographs shown in fig.6 and fig.7 have been taken in the Oct. 1998 by IRS-IB & in May 1999 by IRS-ID respectively, in the same spectral bands. The FCC taken by IRS-1D is based on the data collected by multispectral LISS III sensor. It is relatively high resolution image .It has been used here to see the temporal variation in the area. These two images are of same area, Vaishli district in Bihar, India, and is shown in fig 8 (from www.mapsindia.com) but are taken in different season. Thus by comparing the two images several land use/land cover showing temporal variation like permanent water logging area, temporary water logging area etc have been detected. The identified land use/ land cover areas have also been presented in fig.9.No evidence of snow and hilly area has been found. In the course of interpretation the user can learn the visual image interpretation skill by going through the rules used by the system. The result of the system is satisfactory at the macroscopic level.
Fig. 5: Standard FCC image taken by Satellite IRS - 1D
Conclusion & future work
It may be concluded that the knowledge acquisition, analysis and their structuring to develop knowledge-based rule is major work in this presentatation. Testing the knowledge base and do required correction is relatively smaller part of the work. The system is capable in doing classification at the first level of USGS land use & land cover classification scheme. The developed system is promising and encouraging. It helps in illustrating the major concepts of visual image interpretation and similar system development. The system can be used by the beginners to learn visual image interpretation of the FCC image. The knowledge base of the system stores rule in first come first serve sequence. This is not very strong resolution for conflicts that often arises in the FCC image interpretation. However, by assigning some priority value to each rule in compliance with the human expert’s view stronger conflict resolution can be achieved, as it is independent of ordering of rules. Also the picture base of the system is required to have more interpreted images to make the system useful for lay users. In order to make system more useful it is recommended to integrate digital techniques also. As in the coming decades the need of on screen interpretation of high-resolution satellite image is likely to increase [16], the knowledge base for high-resolution image interpretation can be developed in the same way.
Fig. 6: Location of Vaishali district in Bihar
Acknowledgement
This work was supported by CSIR, NewDelhi. We would also like to thanks to Bihar Remote Sensing Center Patna, for technical support and advice.
Fig. 7: Representation of identified Land use & Land covers (for fig. 4)
References
The science and art of remote sensing has emerged as one of the most important application of space technology in obtaining environmental and natural resources data related to earth using various aerial platforms. These data are being used in large number of areas e.g. agriculture, forestry, archeology, geography, geology, oceanography, ecology etc to obtain useful information for the future planning, management and sustainable development of the natural resources. Such data are collected using different aerial platforms. However, the satellite based remote sensing has become very popular and different countries have launched remote-sensing satellites, for this purpose. The collected data are processed and interpreted in different forms and formats using digital techniques or optical techniques to extract useful information. One of the most widely used data format for information extraction is the infrared False Color Composite (FCC) image. The extraction of information from such images about ground reality is done by image interpretation for which generally three methods namely photo interpretation, spectral analysis and data integration are used. However, for the public, satellite images and the information they bear seem vague and far beyond the understanding of a layman. In the coming decades remote sensing will be one of the key tool for making critical decisions affecting the earth, its environment and resources. So it is important that the informed citizens should have a basic understanding of these technologies and how they are used. Such awareness in mass will ensure the credibility and faith in the accuracy of information derived from the remote sensor data.
In this paper our main concern is with photo interpretation method of IRS-1B standard FCC and to describe the development of a rule based expert system, using expert system shell ESTA (Expert System shell for Text Animation),(Fig.1) for the beginners to learn the visual interpretation skill for the purpose of land use & land cover mapping.
Fig. 1: Expert system shell to expert system
Expert system and FCC image interpretation
Photo interpretation is the visual interpretation of images based on features tone, pattern, shape, size, shadow, texture and association. Most of the conventional digital image processing techniques is based on color or size or texture or tonal variation of each pixel in the image. In contrast to digital analysis of the images, a human interpreter does not interpret the image pixel by pixel. Instead he or she exploits the aggregate information related to various basic image-features of unknown objects along with his scientific knowledge, general knowledge of the phenomena as well as experience to do classification [1,2]. As a consequence, the interpretation result for land use and land cover produced by a well-trained human interpreter is often less crude than the same obtained using digital techniques [3]. For human interpreter it is easy to interpret natural color image but the interpretation of FCC image becomes difficult and requires adequate training and experience [4]. Also different band combinations of satellite data for three primary colors result in different FCC images, which are suitable for different application. Every application of remote sensing deals with a specific subject or integrated process of different subjects. Thus the process of visual interpretation of wide variety of remotely sensed data is a complex intuitive process of combining evidential information from different sources and subjecting such information to an expert’s knowledge, experience and heuristics at each levels namely detection, identification, analysis, recognition and classification of the process. It calls for analysis of a number of related information by a domain expert. So even with on going advances in digital image processing technique the importance and role of human photo interpreter can’t be denied and it is required to train human resources in this skill. Unfortunately, this art has not been taught systematically in the related educational institutions from the last 20 years. As a result this art is being lost [5]. However, in future the scope of on-screen interpretation of high resolution remote sensing data will increase and image identification system providing way to combine together human interpreter and machine interpretation will be required as the maximum interpretation accuracy achieved using digital image processing technique has been reported up to 70 to 80% [6]. The associated information and the logical reasoning that are used by a well-trained human interpreter can be encoded in the from of rules and facts to create knowledge based systems. The need of expert level performance in the remote sensing image interpretation has brought in a shift from domain independent statistical methods to domain specific knowledge based techniques [7]. Thus the activity of image interpretation has similarity with the nature of explorative and qualitative reasoning in the line of artificial intelligence and expert system. Since the interpretation of each FCC images requires different skill and experience so for a human interpreter it becomes difficult to manage properly application of huge knowledge while making decision in photo interpretation. Also the human interpreter may be absent or not easily available. The interpretation by a human is also effected by his mental status like biases, anxieties, fatigue, repeatability of task etc. So the need of computer based support is obvious. The knowledge used in photo interpretation can be represented in logical paradigm that makes prolog [8] based expert system shell ESTA suitable for expert system development. Lot of image interpretation system has been developed in the past. These system are different in number of ways like the knowledge representation scheme, control strategy, application area etc. Most of them are based on spectral features of the image. Very few systems have been developed for developing visual interpretation skill. In the field of land use /land categorization using satellite data a rule based expert system [9] was developed using visual interpretation keys developed by NRSA (National Remote Sensing Agency), India. In other development an expert system for “Land Form Study” has been created, in the Prolog language, for landform type identification and classification in automated and user interactive way [10]. Majority of systems have been developed for automatic image interpretation e.g. the system SPAM [11] for aerial image interpretation is based on production rules. The system MESSIE [12] has used production rules as well as semantic nets. In AIDA system [13] semantic nets and rules has been used to represent knowledge base while ERNEST [14] uses semantic net only. SIGMA [15] contains three expert modules based on frame and rule representation. Recently Jenson et al (16) in his neural network based photo interpretation approach have emphasized use of hybrid approach i.e. visual interpretation and digital interpretation. However, in the neural network based system the classification rules are hidden in its weight and is not useful for a user for learning point of view.
This paper is organized as follows. In the following section and its subjection the proposed expert system development process using ESTA has been described. In the section 4, consultation with the system and result has been presented using standard FCC images produced by IRS-1B & 1D satellites data. The last section contains conclusion and future work.
Fig. 2: Section tree of the system
Expert System Development in ESTA
ESTA
Expert system came out as the first commercial fruit of AI. These are domain dependent computer programme that use knowledge and inference procedure of domain specific human expert to mimic his/her problem solving capability and way in which they solve [17]. In the beginning the expert system development was mainly carried out by the AI people using sophisticated inference engine and search techniques. However, later it was found that the expert system developed by non-AI (domain expert) people using simple inference engine and search techniques outperformed those developed by AI people. This led the AI people to realize the fact that the quality of knowledge highly influences the quality of expert system, which in turn led to the separation of knowledge module from inference and control strategies [18]. This gave a great leap in the expert system technology and resulted in the development of expert system shell. An expert system in the context of expert system shell may be defined as follows:
EXPERT SYSTEM=KNOWLEDGE BASE + EXPERT SYSTEM SHELL (Control &Inference +User Interface).
The separation of the control and inference parts in the knowledge base is a fundamental feature of an expert system shell. ESTA is also an expert system shell developed by PDC (Prolog Development Center), Denmark [19]. By providing with it a knowledge base ,as shown in fig.1, of a specific domain an expert system can be created. Thus
KNOWLEDGE BASE +ESTA= EXPERT SYSTEM.
Fig. 3: A Screen shot when user seek help during consiltator
ESTA system has several advantages over other expert system shell like CLISP. According to PDC, ESTA is easy to use and a great stand-alone environment for constructing advisory and decision support systems. Building advanced knowledge base with ESTA requires no previous programming experience and it is suitable for many problem domains. ESTA is the perfect tool for structuring of knowledge. It includes explanation facilities for the question being asked and for given advice. Details about ESTA can be found in the reference [19] or PDC homepage. It is very easy to use, fast and requires no programming experience. The knowledge base of the ESTA consists of rules represented in its own syntax. It has all the inbuilt facilities to write the rules that build the knowledge base.
Table 1
| Land Cover/land use | Image Characteristics |
| 1. Settlements | Light Gary clustering with particular patterns for the urban area. There may be brownish maroon patches for in between vegetation. For the rural settlement there occurs no particular patterns of such image Characteristics. |
| 2. Agriculture | Identify rabbi if the month of data acquisitionis January or February or March and color is brown red. For the kharif crops same characteristics in image occur if the image data are acquired in the month of September, October or November.(b) Fallow land Fallow land is identified by light gray color within cropped area (red color).(c) Plantation occurs as brownish maroon patches. |
| 3. Forest (a) Dense forests (b) Degraded forest (c) Forest blank (e) Forest plantation |
Dense forests are identified by dark red color patterns. In the case of degraded forest the dark red colour patterns contain small brown or white patches. The blanks in the forest show creamy patches in the dark red/ background. Forest plantations are identified by dark red colour sign of particular pattern. |
| 4. Waste Land (a) Muddy water logging (b) Clear water logging (c) Temporary water logging (d) Permanent water logging (e) Marshy area water logging (f) Gullied land (g) Land with scrub (h) Land without scrub (i) Sandy area |
Muddy water logging occurs as blackish or deep blue spots while clear water logging area is identified by dark/bright blue patches. Comparing the images of rainy season and out of rainy season identifies temporary and Permanent water logging. Marshy area is recognized as a sign of vegetation (red/pink spots) in the water logged (blackish blue/bright blue) area. Gullied land occurs as white/gray spot. The image of land with scrub contains white patches in the land area. Sandy area is classified as bright white coloration along the course of river. |
| 5. Water bodies (a) River/stream (b) Canal (c) Lake/ Reservoirs (d) Embankments |
River/stream is identified as long non-linear path colored with dark blue/ bright blue line in white background. Canals are identified as line segments sign of water. Lake/reservoirs are identified as patterns along the river. Embankment occurs as light gray structure along the river. |
| 6. Others | Grasslands are identified as uneven appearance characterized by red (light to medium gray tones) snow is identified as white patches on the hills. |
Knowledge Base Development
Palpably as an expert system shell, ESTA for knowledge engineer works as tool to create knowledge base and from user side it is used to consult the created knowledge base as shown in the fig.1.Thus the task of developing expert system using ESTA, domain independent module, is centered at developing a domain specific knowledge base, which is a multi-step process. These steps are problem identification, knowledge acquisition and knowledge representation in accordance with the syntax of ESTA. Problem identification is already contained in the previous sections. The rest two are described below.
Knowledge Acquisition
The knowledge required for present study has been gleaned out from available standard literature [3,4,20] and human experts from remote sensing center, Patna using IRS-1B based FCC images. The theme of such interpretation rules, in very brief, for land cover and land use identification is presented in the table 1
The IRS-1B satellite uses Linear Imaging Self Scanner (LISS) I & II sensors, which are approximately equivalent to Thematic Mappers (TM) bands 1,2,3& 4. . So the same knowledge base can also be used to interpret TM based standard FCC image with little variation
Knowledge Representation
The knowledge base for ESTA consists of simple if- then rules. It is also obvious from above table of visual image interpretation key that this domain is based on the specific rules for specific land use/land cover category, so the rule based approach [21] for knowledge representation will be best suited. There are two major knowledge representations in ESTA namely Section and Parameters. Section is top level of knowledge representation and contains the logical rules that directs ESTA how to solve problem, actions to perform such as giving advice, going to other section, calling to routines etc. The first section is always named as start section. Parameters are used as variable and it decides the flow of control among the sections. ESTA accepts four types of parameters namely Boolean or Logical, Text, Number and Category parameters. These parameters serve different purposes. The Boolean parameter is used when the answer to asked question is either Yes, No, or Unknown. Text parameters are used for text object. Number parameter is used to represent numerical values. Category parameters are used when variable takes more than one values. The value for any of the parameter is calculated from end-user’s response to a question, through other parameters or as a result of application of rules. Any parameter consists of declaration field, type field and number of optional field such as explanation field, rules field, picture field, question field etc.
Accordingly, the knowledge base has been developed here too. Three layers of sections for each land use and land category occurring in table1 have been used. These are shown in the fig2, which represents screen shot of the section tree. The rules in each section are arranged in the first come first serve sequence that decides flow of control among the sections. The firing of a rule in any section is governed by the related parameters. All types of parameters have been used in developing knowledge base as per the design requisition. The section consult again represents end of session and can direct control, on user’s response, such that it become possible to do consultation for all land use & land cover types in sequence in each session, otherwise, it needs to start re-consultation for each type. In order to tackle the uncertainty related with any Boolean parameter (when a user responses as Unknown to a question ) multilevel dialog has been designed so that user become aware with what is being asked by the system. The top level of dialog uses standard scientific terms and on each lower level effort has been made to simplify the terms. The declaration field of each parameter contains proper explanation for HOW type explanation while the explanation field contains the explanation to the question related to parameters. Each parameter contains picture name in the picture field, which displayed during consultation. Pictures are stored in picture database in BMP format. On the picture displayed with question of parameter, the land use land cover type has been specified using hotspot editor. Current version of ESTA allows use of picture in the title, parameter, in advice and in action list. However it has been found essential in, some places, to include pictures in explanation. This has been achieved by directing control flow in the advice and pictures have been used to make easier understanding the question. Such an example is shown in fig 3 where the user selects help option. This drives the system in advice window, as shown in the screen shot of fig.4, where the user gets option to see more interpreted FCC images to develop idea about visual image interpretation. The developed knowledge base has been stored in compiled format to optimize the performance
Fig. 4: A Screen shot in response to fig.3
Consultation and Result
The consultation with the system can be initiated by loading knowledge base in memory. ESTA allows only one knowledge base at a time. As the programme runs ESTA goes in start section and looks for the rules. It asks question related to parameters in ESTA Consult window It executes rules in order of the arrangement. The parameter Land Cover places menu of all the land cover and land use types from where user can choose any one and accordingly, the control is transferred to the related section at the second level. At the second level image interpretation is done. User gets here decision of system in advice window. However, as shown in screen shot of fig. 4, the advice window has been also used to provide help on question using additional figures for aforesaid reasons. The interpretation rules in the sections, at second level, have been placed in first come first serve sequence. These sequences have been logically arranged keeping in view minimizing the situation of conflict, which is very common in image interpretation [22]. However, this does not provide strong strategy to overcome the problem of conflict resolution.
Fig. 4: Standard FCC image taken by IRS - 13
In the present work two FCC images, taken by IRS- 1D & IRS-1B satellites, of Hajipur area of district of Vaishali of Bihar have been used for the purpose of identification of the land use/land cover. The photographs shown in fig.6 and fig.7 have been taken in the Oct. 1998 by IRS-IB & in May 1999 by IRS-ID respectively, in the same spectral bands. The FCC taken by IRS-1D is based on the data collected by multispectral LISS III sensor. It is relatively high resolution image .It has been used here to see the temporal variation in the area. These two images are of same area, Vaishli district in Bihar, India, and is shown in fig 8 (from www.mapsindia.com) but are taken in different season. Thus by comparing the two images several land use/land cover showing temporal variation like permanent water logging area, temporary water logging area etc have been detected. The identified land use/ land cover areas have also been presented in fig.9.No evidence of snow and hilly area has been found. In the course of interpretation the user can learn the visual image interpretation skill by going through the rules used by the system. The result of the system is satisfactory at the macroscopic level.
Fig. 5: Standard FCC image taken by Satellite IRS - 1D
Conclusion & future work
It may be concluded that the knowledge acquisition, analysis and their structuring to develop knowledge-based rule is major work in this presentatation. Testing the knowledge base and do required correction is relatively smaller part of the work. The system is capable in doing classification at the first level of USGS land use & land cover classification scheme. The developed system is promising and encouraging. It helps in illustrating the major concepts of visual image interpretation and similar system development. The system can be used by the beginners to learn visual image interpretation of the FCC image. The knowledge base of the system stores rule in first come first serve sequence. This is not very strong resolution for conflicts that often arises in the FCC image interpretation. However, by assigning some priority value to each rule in compliance with the human expert’s view stronger conflict resolution can be achieved, as it is independent of ordering of rules. Also the picture base of the system is required to have more interpreted images to make the system useful for lay users. In order to make system more useful it is recommended to integrate digital techniques also. As in the coming decades the need of on screen interpretation of high-resolution satellite image is likely to increase [16], the knowledge base for high-resolution image interpretation can be developed in the same way.
Fig. 6: Location of Vaishali district in Bihar
Acknowledgement
This work was supported by CSIR, NewDelhi. We would also like to thanks to Bihar Remote Sensing Center Patna, for technical support and advice.
Fig. 7: Representation of identified Land use & Land covers (for fig. 4)
References
- Estes, J. E., Hajic, E. J. and L. R. Tinney, (1983). Chapter 24:Fundamentals of Image Analysis: Analysis of Visible and Thermal Infrared Data, Manual of Remote Sensing, American Society for Photogrammetry & Remote Sensing, Bethesda, pp. 987-1124. (1)
- Jensen J. R.( 2000). Remote Sensing of The Environment: An Earth Resource Perspective, Prentice-Hall, Inc., Upper Saddle River, NJ, 544 p. (2)
- Philipson, W. (1997). Manual of Photographic Interpretation. American Society for Photogrammetry & Remote Sensing, Bethesda, 830 p. (3)
- Short, N.M., (1999). The Remote Sensing Tutorial, NASA/Goddard Space Flight Center. On line http://rst.gcfc.nasa.gov/Front/tofc.html (4)
- Estes, J. E. and J. R. Jensen, (1998). Chapter 10: Development of Remote Sensing Digital Image Processing and Raster GIS, The History of Geographic Information Systems: Perspectives from the Pioneers, Prentice-Hall, Inc., Saddle river, NJ, pp. 163-180. (5)
- Jensen, J. R., (1996). Introductory Digital Image Processing: A Remote Sensing Perspective, Prentice-Hall Inc., Upper Saddle River, NJ, 318 p. (6)
- Deekshatulu, B.L.(1991) “Management of Natural Resources -Use of AI,” In AI and Expert System Technology in Indian Context”, Ed. By V.V. Sharma, Deekshatulu, Vishwandhan, TMH, New Delhi. (7)
- Kowalaski R.A., 1982" Prolog as Logic Programming Language,” Proceedings of AICA Congress, Pavia, Italy. (8)
- Ramani, K. & Gautum, N.C., 1990 “ Expert System, for Identification of Land Use/Land Cover Categories from Satellite Data”, Proceedings of the Seminar on AI, Expert System & Knowledge Based System, ISRO Hqts., Bangalore, India. (9)
- Rao, D.P. (1991). Management of Natural Resources using Expert System Preliminary Result of Some Case Study, Expert system Technology in Indian Context Ed. by V.V. Sharma, Deekshatulu, Vishwandhan, TMH, New Delhi. (10)
- McKeown, D., Wilson, A. & McDermott, J. (1985). Rule-Based Interpretation of Aerial Imagery. IEEE Trans. on Pattern Analysis and Machine Intelligence 7(5): 570{585. (11)
- Clement, V., Giraudon, G., Houzelle, S. & Sadakly, F. (1993). Interpretation of Remotely Sensed Images in a Context of Multisensor Fusion Using a Multispecialist Architecture.IEEE Trans. on Geoscience and Remote Sensing 31(4): 779{791. (12)
- Liedtke, C.-E., Buckner, J., Grau, O., Growe, S. & Tonjes, R. (1997). AIDA: A Systemfor the Knowledge Based Interpretation of Remote Sensing Data. 3rd Int. Airborne Remote Sensing Conference and Exhibition, 313{320, Copenhagen, Denmark. (13)
- Niemann, H., Sagerer, G., Schroder, S. & Kummert, F. (1990). ERNEST: A semanticNetwork System for Pattern Understanding. IEEE Trans. on Pattern Analysis andMachine Intelligence 12(9): 883{905. (14)
- Matsuyama, T. & Hwang, V.S.-S. (1990). SIGMA: A Knowledge-Based Aerial Image Understanding System. New York: Plenum Press, p. 277. (15)
- Jensen J. R., F. Qiu and Patterson K, (2001), A Neural Network Image nterpretation System to Extract Rural and Urban Land Use and Land Cover Information from Remote Sensor Data, Geocarto International, Vol 16,No.1, Hong Kong. (16)
- Luger, George, F.& Stubblifield, A.(1983), “Artificial Intelligence: Structures and Strategies for Compex Problem Solving. “ The Benjamin/Cummings Publishing Company, Inc. California, 2nd Edition. (17)
- Al-garni, A.M., Al-sari (1994) Remote Sensing Geology and Expert System, ASPRS/ACSM, Annual Convention & Exposition, Baltimore, Vol.1, pp47-59. (18)
- Introduction to ESTA (Expert System Shell for Text Animation), Prolog Development Center, Denmark. (19)
- Thomas, M, & RALPH, W. (1999), “Remote Sensing and Image Interpretation” (20)
- John Wiley and Sons, 4th edition, New York.
- Leung, Y And Leung, K.S. (1993) “ An Intelligent Export System Shell for Knowledge Based Geographical Information System 2, Some Applications, International Journal of Geographical Information System. (21)
- Pakiarajah, V., Crowther, P., & Hartent, J., Conflict Resolution Techniques for Expert Systems Used to Classify Remotely Sensed Satellite Images, 5th International Conference on GeoComputation, UK. (22)