An Intelligent Environmental Monitoring System

The Need for Environmental Monitoring

Changes in the environment can be caused by either natural catastrophies like floods, typhoons, and volcanic eruptions, or the human factor like industrial pollution and urban development. Whatever the cause, such changes may result not only in the loss of life and property in the short term, but also environmental degradation in the long term.

To conserve the environment, it is essential that environmental changes such as variations in the conditions, patterns, and dynamics of air, water, and land resources, are effectively monitored.

Recent advancement in Geographic Information Systems (GIS) technology has enabled environmental scientists to incorporate spatial data collected via diverse sources for monitoring purposes. In the GIS can be found digitized maps, demographic figures, and multitemporal information about the temperature, humidity, size, boundary, etc. of specific locations.

Closely linked with GIS is remote sensing technology under which special data are acquired in the form of images depicting spatial variation of terrestrial objects in terms of spectral reflectance. Remotely-sensed images range from the conventional aerial photographs taken from airplanes to the more sophisticated multispectral and multi-resolution digital data acquired by satellite sensors.

At present, both GIS data and remotely-sensed images can be stored, managed, processed, and retrieved in an integrated geographic information system (IGIS). But if scientists want to obtain high-level analysis and sophisticated interpretation of these two sets of data together, the IGIS has first to be upgraded and made more `intelligent'.

Drs. Fung Tung and Leung Yee of the Department of Geography and Dr. Leung Kwong Sak of the Department of Computer Science are working on such a new, more intelligent IGIS, and have been awarded a grant of $800,000 by the Research Grants Council for the purpose.

The project involves image processing, expert knowledge, GIS, databases, as well as other computing algorithms.

First Phase: To Improve Remote Sensing Image Processing Systems

Conventional remote sensing image processing systems are developed on the basis of statistical pattern recognition and image analysis methodologies. Techniques have been developed to enable the systems to detect changes and classify images, but such techniques are very basic and their results are crude when compared with those of a skilled photointerpreter.

A good photointerpreter can efficiently utilize the image characteristics of tone, colour, shape, size, texture, pattern, shadow, and site, and synthesize them with his/her knowledge of a specific region or discipline to produce detailed and accurate interpretations of environmental changes.

Researchers at CUHK have set out to formalize and automate the manner in which a photointerpreter performs his or her work, and to produce an intelligent image processing system that could provide high-level analyses and interpretation of remotely-sensed data.

Second Phase: To Build a Common Platform to Integrate Subsystems

This intelligent image-processing system will then be linked with GIS and an expert system shell to form a revitalized IGIS that is more powerful and effective in monitoring environmental changes and solving complex problems.

The expert system shell was first designed by Drs. Leung Yee and Leung Kwong Sak back in 1990. It has since been improved and is now able to imitate the decision-making processes of the human brain, which has the ability to summarize data and focus on decision-relevant information. As long as relevant GIS and remotely-sensed data are fed into the system, solutions will be supplied to specific problems.

To achieve the integration of the subsystems, novel interfaces and overall control algorithms will be designed. They will be implemented and tested individually first, before the integrated system undergoes further tests to ascertain its effectiveness.

Third Phase: Case Study and Evaluation

The next step will be to use GIS and remotely-sensed data about

urban development in Hong Kong to evaluate the efficiency and reliability of the new IGIS. Efficiency is tested in terms of the computing time it takes to generate results. Reliability is tested in terms of the accuracy of information produced.

The research project attempts to improve conventional low-level image analysis methods by incorporating a high-level knowledge-based technology, bring about automation for image interpretation, increase information accuracy with an intelligent expert system, and build up local expertise and databases for purposes of application in Hong Kong. What results should be an extremely powerful research tool that can contribute significantly to not only environmental studies in the territory, but its social and economic growth as well.