Overview

The objectives of this project are: (i) to establish an Energy and Environmental Information Resources (EE-IR) Center; (ii) to conduct research work in Internet computing in support of the EE-IR Center; and (iii) to enhance the facilities of the NASA/USL Regional Applications Center. Through this work, we have collaborated with DOE's Office of Scientific and Technical Information (OSTI), and with the U.S. Geological Survey's National Wetlands Research Center. In May 1998, Vijay Raghavan (Project Co-manager), Marie Erie (Doctoral Student), and Michelle LeBlanc (Doctoral Student) attended OSTI's 1998 InForum Meeting and presented a paper describing their work. Following the meeting, Vijay Raghavan and Henry Chu (Faculty Investigator) visited OSTI and discussed research directions with OSTI's scientists. In the following, we describe our progress in more detail.

Energy and Environmental Information Resources Center Development

The EE-IR Center is developed in collaboration with the National Wetlands Research Center (NWRC), and with OSTI. In December 1997 and January 1998, we had two meetings, and identified and interacted with potential end-users of the EE-IR Center and the GIS community. While we are working with NWRC on data collection and ingest issues, we are also developing software tools addressing different aspects of digital library development. We have studied the strategies for enhancing pre-existing Web documents in order that just the relevant portions (or components) of a document can be retrieved. We have also developed an approach for searching product catalogs on the Internet, based on concepts, so that an intermediary can conduct the search on different databases. Our approach to concept-based retrieval should be of interest to refining certain kinds of canned searches that are often made in the DOE Information Bridge.

In April 1998 Dan Foley (metadata librarian), Adam Chandler (systems librarian), and Suzanne Harrison (research associate) started working at the NWRC on data collection for the EE-IR Center. The research work at the Center is to locate data in numeric, text, and graphical display and to document the data in a standard Federal Geographic Data Committee (FGDC) or National Biological Information Infrastructure (NBII) format. The metadata will be served on the Internet in standard FGDC or NBII formats. In addition, the data will be mapped to Machine-Readable Cataloging (MARC) to include in the data server at the center and an international data server, WorldCat. Data served by the EE-IR Center will be added to a library of data using collection development decisions compatible with the scope of the center. Data is to limited to the Gulf Coast area of Louisiana with a central theme of data for contaminants or pollution studies. Data for this geographical area can be broadened to other geographical areas after the strategy and links are documented. Data initially will be documented with Dublin Core elements and these mapped to MARC fields and NBII metadata standards. Another 1998 research task is to enhance the Web to Database (WDB) software that allows the linking of relational databases to a Web browser.

Research Work

Our research work is organized under the following four headings: multimedia retrieval; data mining; parallel and distributed computing; media technology and data visualization.

In multimedia retrieval, we have completed a project in description and evaluation of an adaptive image retrieval system called Web-IDBS. We have made a qualitative comparison of different client-server implementations that used the Java programming language for an adaptive image retrieval application. We have developed strategies for improving the efficiency of feature-based (color, texture or shape) image retrieval.

In data mining, we are investigating the impact of data mining on database security. Specifically, we are focusing on evaluating a protected data element's risk of disclosure in the context of classification learning. We partitioned classification methods into two categories, and developed evaluation algorithms for the two categories. We are currently working on experimental investigations of these algorithms. One of the investigators has guest edited a special issue of the Journal of American Society for Information Science (April 1998, Wiley) on ``Knowledge Discovery and Database Mining". One of the papers in that special issue reports our research results on efficient algorithms for feature selection, in the context of designing classifiers based on the theory of rough sets.

In parallel and distributed processing, we investigated the design of a high-performance server cluster for web applications. A cost effective approach to setting up a web server is to incrementally scale up the hardware as web access traffic increases. Workstation clusters are an attractive solution for such applications. The performance metric in evaluating such solutions is the average waiting time of a task to be completed. As the workload of one workstation reaches a threshold, it can send some of its tasks to its proxies, who are considered to be logical neighbors. We developed such load balancing algorithms for two different network configurations, and evaluated the performances empirically on actual networks in our department. The preliminary results are that the load balancing algorithms provide substantial improvements in average waiting times, especially at high traffic densities.

In media technology, we investigate issues related to transmitting imaging and video data over the Internet. Specifically, we developed a method for progressive transmission of image data so that the rendered image is perceptually lossless. In view of the large number of images that are printed after being downloaded, we incorporated halftoning into the compression method. Video data convey information in the most vivid form, albeit at a cost of substantial bandwidth requirement. For the forseable future, the Internet remains a low bit rate channel for video transmission, especially when compared to other channels. We investigated the use of wavelet transform, a method that has proven to be effective for low bit rate image transmission, for low bit rate video compression. Our method use a pair of wavelet-based processing modules as wrappers around an international standard codec (the ITU-T H.263). We showed that we can have better video quality (in terms of still frame and motion) at fixed bit rates compared to the standard codec.

In data visualization, we developed a new color halftoning algorithm for indexed color display. Compared to the standard error diffusion algorithm, our new method has fewer color spikes and no color shifts. We also developed a method for visualizing scalar fields on three-dimensional surfaces. The application of our method is in visualizing EEG-derived cortical potential fields on MRI-derived cortical surfaces---a method that fuses two brain imaging modalities.

NASA/USL Regional Applications Center Enhancement

We acquired a GOES geostationary satellite receiver to augment our data collection by the AVHRR receiver. We improved the computing environment by acquiring color workstations and several servers. The NASA/USL RAC is one source of information for our EE-IR Center. One of our ongoing projects is to derive and display the flow of water currents in the Gulf of Mexico from AVHRR images. Another ongoing project involves the development of tools for the integrated use of geographically referenced datasets and for the mining of these data sets to aid in decision making relative to certain biological or environmental issues. The approach involves the re-engineering of applications currently implemented using Geographic Information Systems so that they use state-of-the-art technologies (Java, CORBA, CGI scripts, etc.).

Research Publications and Human Resources Development  

Journal articles- 3; Conference articles- 3, Ph.D. dissertations- 2, Ph.D. prospectus- 1; M.S. theses- 2; M.S. Project- 1; Technical Reports- 2.