Home     Agenda     Registration     FAQ     Abstracts  

Speaker Abstracts/Position Statements

Multi-modal medical volume reconstruction, analysis, and visualization
Shadi Ashnai, Sang-Chul Lee, Peter Bajcsy - NCSA
Kenneth L. Watkin - College of Applied Health Science
Monday, May 21
6:30pm - 8:00pm

This poster presents the problem of integrating, analyzing and visualizing multi-modal brain data. The multi-modal images are represented by magnetic resonance images (MRI), diffusion tensor images (DTI), and bright field microscopy images of histological cross sections of the specimens stained with the Nissl stain. While the MRI and DTI data are acquired directly as 3D volumes, 3D histological volume has to be reconstructed from multiple high spatial resolution 2D cross sections. The three 3D volumes of different modalities are spatially integrated into the same coordinate system using computer assisted techniques. The major challenges of the multi-modal integration are that (a) the spatial resolutions of these volumes differ in every dimension, (b) the appearance of the same physical tissue varies across modalities, (c) the modality specific measurements represent grayscale (MRI) or color (Histo) or vector (DTI) values, and (d) the file size of 3D volumes requires significant computational resource and scalable algorithms. Our approach is based on performing computer assisted, coarse to fine spatial resolution, integration with the user support by 3D visualization and quantitative feedback on the estimated integration accuracy. We demonstrate such integrations, clustering of line segments (fibers) in DTI and extraction of foreground in MRI and Histo volumes, and finally 3D visualization of multi-modal information. Our work is applied to understanding gender specific and stuttering characteristic patterns in animal brain models.

back to the agenda



Computer Assisted Extraction of Information and Knowledge from Images and Other Data
Peter Bajcsy - NCSA
Wednesday, May 23
9:00am - 9:30am

This talk will present several examples of computer assisted learning from either images only or images fused with other data. The examples come from multiple scientific and industrial domains that would like to use the exponentially growing amounts of raw image and other data with smart algorithms and high-performance computing resources. The talk will outline a selected set of problems in geospatial-, bio-, medical-, document- and sensor-informatics domains and describe a set of freely available prototype solutions.

back to the agenda



Real Application Performance Analysis on NCSA Systems
Gregory Bauer, Nahil Sobh - NCSA
Monday, May 21
6:30pm - 8:00pm

To reduce turn-around time and memory usage of your application, and to maximize your system's workload, it is essential to understand your application's performance with respect to the system's architecture. We will discuss our experience in analyzing the performance of some real applications using tools that are available to the users of NCSA systems. With performance opportunities in mind, the "best practices" for memory use, communication and I/O will also be discussed.

back to the agenda



Image analyses of very large size collections of scanned documents
Melvin Casares, Kirk Hard, Peter Bajcsy - NCSA
Daniel W. Stowell, Stacy McDermott - Abraham Lincoln Presidential Library
Monday, May 21
6:30pm - 8:00pm

This poster describes the challenges of automatic classification and cropping of very large size collections of scanned paper documents. The papers represent a large collection of incoming and outgoing correspondence of Abraham Lincoln, and are scanned at the National Archives and the Library of Congress in Washington, DC. The ultimate goal of these images is to provide an online research and reference work that will provide the color images, transcriptions, and editorial matter for these historical documents, including Lincoln's extensive legal practice. In our work, we design algorithms for automatic classification of images containing the documents with or without additional background items. Then, the images with background items are automatically analyzed to identify the crop region containing only the documents. Finally, the algorithms are applied to a large volume of images that will consist of 200,000 to 300,000 pages with each page equal to about 150MB. The cropped images will be available on-line via the Illinois Historic Preservation Agency and the Abraham Lincoln Presidential Library and Museum in the future.

back to the agenda



Designing scientific information and communicating to non-expert audiences
Donna Cox - NCSA
Tuesday, May 22
1:30pm - 2:15pm

Donna J. Cox will present a visual feast of digital data-driven scientific visualizations from her collaborative work at NCSA. She coined the term "Renaissance Teams" in 1985 to describe interdisciplinary research teams focused on challenges in scientific visualization. In the early years, teams were small local groups, but now through grid technologies, many teams are distributed remotely and work collaboratively through global technologies. She collaborates, designs and aesthetically directs data-driven visualizations of supercomputer simulations from a variety of disciplines ranging from oceanography to astrophysics. She demonstrates both rigorous design methodologies and works with educators and focus groups to provide solutions to the challenging work of data-driven visualizations for non-expert audiences. She and her team have developed new visualization methods and virtual reality tools that provide novel ways for researchers to collaborate over remote distances and high-speed networks. Cox will describe these technologies and the resulting high-resolution, high-fidelity digital scientific animations for museums and high-definition television.

Technology is exponentially transforming the way that people collaborate, scientists 'see', and designers invent. Cox will present a voyage of how art and science have converged through computer graphics technology. Her greatest passion is to employ new technologies to inspire general audiences with the awe and mystery of science. This presentation will demonstrate advanced graphical techniques in the making of 'high art' science for insight and public presentation.

back to the agenda



Environmental Cyberinfrastructure Demonstration
Steve Downey - NCSA
Monday, May 21
6:30pm - 8:00pm

The CyberCollaboratory is a collaborative space where communities of researchers, practitioners, and policy-makers, and others come together to share knowledge and information, analyze data, solve problems, and collaborate on publications. The CLEANER Project Office is using the CyberCollaboratory to support over 100 researchers and educators engaged in WATERS Network planning.

back to the agenda



Increasing Productivity with Windows Compute Cluster Server 2003
Kyril Faenov - Director of Windows High Performance Computing, Microsoft
Tuesday, May 22
9:00am - 9:45am

HPC has undergone a revolution with the advent of commodity compute clusters. Where once HPC was the domain of a specialized few researchers, now virtually any enterprise can gain access to large aggregates of processing power holding out the promise of truly personal supercomputers. Yet building, managing, programming and using these clusters can be a major barrier to entry for compute-power hungry users. The hardware and software are complex and often are controlled by different vendors. To make this revolution complete we need to reach out to organizations and users with an environment that is familiar and easy to use and exploit. This is where Microsoft products, user interfaces, and developer tools add value, by integrating these specialized resources into an environment that allows seamless computing, communication, collaboration and ultimately creativity and productivity.

Last year Microsoft introduced Windows Compute Cluster Server 2003 (CCS), cluster software that can be easily and quickly deployed using standard Windows deployment technologies, featuring the Microsoft Message Passing Interface (fully compatible with MPICH2), integration with Active Directory enabling role-based security for administration and users, and the use of Microsoft Management Console to provide a familiar administrative and scheduling interface. Windows CCS is designed to fit easily into an organization's IT infrastructure while integrating into the user's desktop, productivity and collaboration tools. The attainment of this goal has been demonstrated at a rapidly growing number of customer sites where Windows CCS has boosted user productivity and reduced time spent on administration.

back to the agenda



CluMon: Cluster Monitor
Joshi Fullop - NCSA
Monday, May 21
6:30pm - 8:00pm

NCSA's scalable and fault tolerant solution to the ever-growing cluster monitoring needs. CluMon is currently capable of collecting data from thousands of compute nodes and offering correlated views of that information users via a Web interface as well as a Web-based API. With a plugin architecture, CluMon can be expanded to deal with most any scheduler or data source for a complete, customized monitoring solution.

back to the agenda



Designer Viscoelastic Materials Tailored to Minimize Failure Probabilities and Prolong Lifetimes
Harry H. Hilton - NCSA
Monday, May 21
6:30pm - 8:00pm

Viscoelastic materials are known for their ability to dissipate energy. This property has been successfully used by the author and his colleagues to produce effective passive structural control for column and plate creep buckling, various vibratory modes, and aero-viscoelastic phenomena, such as torsional divergence, lifting surface and panel flutter, and attenuation of aerodynamic noise in panels. In self-excited systems the application of increased dissipation may stabilize or destabilize such systems depending on the influence of damping and all other forces on phase relations.

Conventional design and analysis formulations call for use of the best off-the-shelf material. In these research projects, relaxation moduli are tailored through prescriptions of appropriate functionally graded viscoelastic materials to produce the desired designer material properties and performance. Relaxation moduli are, of course, highly temperature sensitive and performances are evaluated relative to operational demands.

Detailed analyses and simulations have been performed to establish the efficacy of these designer protocols in the presence of different relaxation functions and loading conditions.

The ultimate "proof of the pudding" is found in the selection of optimal designer material properties and their spatial distributions that lead to low failure probabilities and long survival times. Research results have applications to flight vehicles (airplanes, missiles, space vehicles), ground transportation (automobiles, railroads, cargo containers), submarines, electronic packaging.

back to the agenda



Real-time 3D object detection and tracking using thermal and visible imagery for tele-immersive environments
Miles Johnson, Peter Bajcsy - NCSA
Monday, May 21
6:30pm - 8:00pm

Automatic real-time object detection in complex social environments remains a significant challenge. The need for 3D object detection and tracking information arises in a wide range of application domains, from tele-immersive systems to hazard-aware spaces. Our research approaches this problem through integrating 2D thermal imagery with a visible stereoscopic camera system. By recursively estimating dense depth of thermal pixels, we are able to re-project thermal information into the visible camera frame, which is necessary for pixel level information fusion. The addition of thermal infrared information provides multiple benefits to current visible-wavelength systems: (a) a very robust and efficient method of extracting objects of interest from the background in an illumination independent manner, and (b) the ability to distinguish real objects from displayed or projected images. In our presentation will show that these benefits not only lead to improved stereopsis results, but also add an enhanced ability to interpret scenes involving complex tele-immersive environments for everybody built at NCSA and other universities connected to our environment.

back to the agenda



A Domain Decomposition Methodology for a Combined Structured and Unstructured Meshes
Rooh-ul-Amin Khurram, Nahil Sobh - NCSA
Monday, May 21
6:30pm - 8:00pm

Combined structured and unstructured meshes, or composite meshes, are often employed to accurately simulate physical phenomena over complex geometry. Standard domain decomposition technique are not suitable in partitioning composite meshes. In this poster we will report on our experience in developing a domain decomposition strategy that combines the strength of both unstructured and the structured domain decomposition methods.

back to the agenda



A Methodology for Solving Fluid-Structure Interaction Problems
Rooh-ul-Amin Khurram, Nahil Sobh - NCSA
Monday, May 21
6:30pm - 8:00pm

As CFD tools are becoming more popular and widespread in engineering analysis and design, there is a growing trend toward applying these methods to analyze and understand complex problems. Simulation of fluid-structure interaction problems is now possible due to the availability of powerful computers and advanced methods. These advances have enabled engineers to design improved products within a short turnaround time. Moving mesh techniques is the key technology that had enabled the formulation of an accurate approximation to the fluid-structure interaction problems. Rooh-ul-Amin Khurram will report on the mesh moving technique that is integrated into the physical formulation. A number of examples will be shown to verify the accuracy of the method.

back to the agenda



Image pattern recognition and data mining applied to groundwater recharge and discharge (R/D) rate models
Chulyun Kim, Peter Bajcsy - NCSA
Yu-Feng Lin (Forrest) - Illinois State Water Survey
Monday, May 21
6:30pm - 8:00pm

Groundwater recharge and discharge (R/D) rates and patterns are difficult to characterize, and currently no single method is able to estimate R/D rates and patterns for all practical applications. Therefore, cross analyzing results from various estimation methods and related field information is likely to be superior than using only a single estimation method. We have developed a geographic information system package, called the Pattern Recognition Utility (PRU), to help hydro-geologists estimate R/D in a more efficient way than conventional methods. The PRU uses numerical methods and image processing algorithms to estimate and visualize R/D patterns and rates. The PRU includes a default R/D estimation code using a finite difference mass balance approach in 2D. This default R/D estimate code only requires data for water table, bedrock elevations and hydraulic conductivities instead of labor intensive and time consuming field R/D measurements. In order to obtain more accurate R/D rates at locations without direct or indirect measurements, we developed a data-driven verification method in the context of auxiliary information such as land coverage, soil type, topographic maps and previous estimates. The verification method is based on a decision tree algorithm and the Minimum Description Length (MDL) principle. The outcome of our research is not only a new verification method but also a set of possible rules using auxiliary information governing the R/D pattern.

back to the agenda



PDF document cleansing and automated information extraction
Rob Kooper, Sang-Chul Lee, Chulyun Kim, Peter Bajcsy - NCSA
Monday, May 21
6:30pm - 8:00pm

We present a set of methods for cleansing and parsing out objects contained in Portable Document Format (PDF). The methods described focus on text and image PDF objects. In order to allow machine understanding of the PDF contents, it is necessary to associate PDF objects that are supposed to convey certain information via rendering of a PDF document. For example, we have identified cases of fragmented words, duplicated characters, and broken lines or paragraphs into subsets. Similarly, we found images being contained within each other or partially overlapped and single images being chopped into smaller pieces. We have prototyped methods for cleansing these cases that enable automation of information extraction. We have tested the methods with PDF documents of newspaper ads since the documents contain many images and textual captions of interest. The contribution of our work is in demonstrating the cleansing problems, classifying the image and text pre-processing cases, and designing several cleansing methods for statistical evaluation of PDF document and loader quality, as well as for preparing the PDF documents for automated information extraction.

back to the agenda



Extraction of handwritten information from paper forms
Sang-Chul Lee, Tim Nee, Radha Nandkumar, Peter Bajcsy - NCSA
Weimo Zhu, Heidi Krahling - Department of Kinesiology
Monday, May 21
6:30pm - 8:00pm

We consider a problem of automated information extraction from hand-written forms, specifically used by the Department of Human Service (DHS), Illinois. The process includes sectioning and re-collection of the printed forms for traditional hand-written filing, raw image acquisition of the collected forms by scanning, and semi- and fully-automated information extraction for future data mining support. The automated data extraction from a hand-filled form is considered as a hard problem unless a standardized form is used, e.g., teleform. For information extraction without modifying the existing form, we prototyped automated image processing techniques such as robust registration of geometrically distorted input images, segmentation of the input fields, and classification of binary answers with high confidence. The processed data may be stored in a database, or as a pdf form with annotations, or a plain text form associated with raw images.

back to the agenda



Data integration and information gathering about decision processes using geospatial electronic records
Sang-Chul Lee, Rob Kooper, Pater Bajcsy - NCSA
Monday, May 21
6:30pm - 8:00pm

The size, complexity and heterogeneity of geospatial collections pose formidable challenges for the National Archives in terms of high performance data storage, access, integration, analysis and visualization. These challenges become even more eminent when temporal aspects of data processing are involved as in the context of high assurance decision making and high confidence application scenarios. In this work, we present analyses of several challenges related to preserving and reconstructing computer assisted decision processes. The challenges include definitions and selections of information granularity, designs of information gathering and reconstruction mechanisms, evaluations of reconstruction value, and understanding of computational costs associated with preservation and reconstruction of computer assisted decision processes. The ultimate goal of our research is to understand the cost of long term preservation of electronic records using the cutting edge technologies, high performance computing and novel computer architectures. Our approach has been to design and prototype a simulation framework for archivists that incorporates multi-disciplinary knowledge and provides a platform for testing hypotheses to meet the future archival needs in preserving decision processes using electronic records. For this purpose, we have prototyped a simulation system called "Image Provenance To Learn" (IP2Learn) that is designed for a class of decisions based on image inspection. It allows users to analyze computational costs of information gathering as a function of information granularity and then assess the potential value of preserved information from decision process reconstructions. We describe the simulation system components and illustrate how IP2Learn simulations could improve our understanding of future archival needs.

back to the agenda



Water quality monitoring data integration and modeling for pollutant load estimation
Qi Li, Peter Bajcsy - NCSA
Momcilo Markus - Illinois State Water Survey
Monday, May 21
6:30pm - 8:00pm

Although nutrients are essential to aquatic ecosystems, excessive amounts of nutrients can be very harmful. High levels of nutrients are often a result of non-point source pollution, a diffuse source of pollution that cannot be attributed to a clearly identifiable, specific physical location, such as the nutrients that enter rivers and streams and originate from any land use, or from air pollution. Non-point source pollution, in particular nutrients and sediment pollution, has been identified as a major cause for water quality impairment in the United States. Various programs and management practices have been designed to reduce the amounts of pollutants, and thus, to better protect water quality in the rivers and streams. A cost-effective water quality monitoring has a key role in quantifying the trends in water quality and the degree of water quality impairment. Loads, the mass of nutrients or sediment transported by a stream during a given period of time, are important in water management, estimating total maximum daily loads, and in prioritization and evaluation of best-management practices.

This poster presents a software prototype for water quality monitoring data integration and modeling for purposes of pollutant load estimation. Scientists have been relying on the historical water quality monitoring data to understand river and watershed pollution. However, there is unbalanced temporal data intensity between stream flow rate and nutrient concentration: while flow rate is typically collected more frequently, concentration for various nutrients is often sampled monthly or even less frequently. To estimate the missing concentration data, domain experts have developed various equations to correlate flow rate to nutrient concentrations, along with multiple correction methods to reduce estimation error. Given the large number of water stations and the long periods with missing data, we developed a software prototype that (a) integrates water quality data collected by multiple agencies, and (b) supports multiple modeling and bias correction techniques. The results of this research will have an important role in predicting water quality parameters, determining the most cost-effective (optimal) sampling design, determining the trends in water quality and evaluating the effectiveness of management practices to protect water quality of our rivers and streams.

back to the agenda



Cyber-integrator: A highly interactive scientific process management environment to support earth observatories
Luigi Marini, Rob Kooper, Peter Bajcsy - NCSA
Monday, May 21
6:30pm - 8:00pm

This poster presents a novel process management environment called CyberIntegrator to support diverse analyses in Earth Observatories. These analyses are very human time consuming and hard to reproduce because of the lack of in-silico scientific process management and because of the diversity of data, software and computational requirements. The motivation for our work comes from the need to build the next generation of in-silico scientific discovery processes that require (a) access to heterogeneous and distributed data and computational resources, (b) integration of heterogeneous software packages, tools and services, (c) formation and execution of complex analytical processing sequences, (d) preservation of traces about scientific analyses and (e) design of secure collaborative Web-based frameworks for sharing information and resources. The goal of the presented work is to describe a modular architecture and key features of a workflow that provides a process management environment for automating science processes, reducing the human time involved and enabling scientific discoveries that would not be possible without supporting software and hardware infrastructure. Our approach to solving the above problem is based on adopting object-oriented software engineering principles, designing a modular software prototype, and focusing on user interfaces that simplify complex analyses including heterogeneous software integration. The main novelties of our work lies in (1) formalizing the software integration framework using object-oriented software engineering principles, (2) designing a browser-based modeling paradigm for step-by-step composition of workflows, (3) enabling provenance to recommendation feedback for workflow auto-completion, and (4) providing capabilities to publish, run, monitor, retrieve, re-use and re-purpose workflows from local and remote computational resources for long-running workflows.

back to the agenda



MAEviz
Terry McLaren - NCSA
Monday, May 21
6:30pm - 8:00pm

MAEviz is a collaborative effort between the Mid-America Earthquake Center and NCSA to develop the next generation of seismic risk assessment software which merges the MAE Centerís research methods and tools to support Consequence-based Risk Management, or CRM. Consequence-based Risk Management is a new paradigm for seismic risk reduction across regions or systems that incorporates point data for all components of seismic risk modeling and quantifies the risk to societal systems and subsystems.

back to the agenda



NCSA, Global Outreach and PRAGMA 13
Radha Nandkumar - NCSA
Wednesday, May 23
9:30am - 10:00am

Radha Nandkumar, director of ICARE, will discuss the collaborations between NCSA and its International Affiliate institutions, which span several continents. One such affiliation is the Pacific Rim Applications and Grid Middleware Assembly (PRAGMA). NCSA will host the thirteenth PRAGMAworkshop, representing a consortium of 30 member institutions from 18 countries, in September 2007. Nandkumar will describe some of the significant collaborative projects in the international arena, how these projects are enabled by various fellowships, and how the private sector partners can participate in them.

back to the agenda



Supercomputing the Universe: Challenges for Data Management in the Large Synoptic Survey Telescope Project
Ray Plante - NCSA
Tuesday, May 22
2:15pm - 3:00pm

The Large Synoptic Survey Telescope will provide unprecedented three-dimensional maps of the mass distribution in the universe, in addition to the traditional images of luminous stars and galaxies. These mass maps can be used to better understand the nature of the newly discovered and utterly mysterious dark energy that is driving the accelerating expansion of the universe. The LSST will also provide a comprehensive census of our solar system, including potentially hazardous asteroids as small as 100 meters in size. Finally, LSST's rapid scans of the sky will open the "time window," searching for faint bursts of light. When operating, LSST will be the primary view of the optical sky for most astronomers around the world (~4,000 science users). It will be one of the most popular (if not the most popular) view of the sky for students and the general public (~10,000 public users).

The raw images (on the order of 15 terabytes per night) must be transmitted from the telescope and processed in real time, generating a total of 130 terabytes of data per night, in order to produce alerts about time-critical discoveries that must be followed up by other telescopes. The demanding data management requirements driven by the high data rate and the need for real-time alerts will require significant advances in cyberinfrastructure. NCSA is partnering with a dozen other academic institutions and Department of Energy labs to design and develop an advanced and highly automated data management system for processing and distributing LSST data for the astronomical community. Our team at NCSA is leading a collaboration to produce a prototype system based on grid technologies which enables high-performance, parallel I/O and robust, automated processing workflows that run on NCSA supercomputers. We are leveraging considerable expertise in both hardware and middleware at NCSA and driving the maturation of general-purpose technologies that can be applied to other grid-based projects.

back to the agenda



Matrix Management Tool
Greg Pluta - NCSA
Monday, May 21
6:30pm - 8:00pm

Matrix Management has been successfully implemented for decades. However, it has been shown that organizations without a clear definition of responsibilities to project and functional managers will not succeed with Matrix Management. This, combined with the demands to run a large IT operation adds another layer of complexity. The Matrix Management Tool (MMT) is a tool allowing managers to better manage such an organization.

back to the agenda



Visualization Pipeline
Dave Semeraro - NCSA
Monday, May 21
6:30pm - 8:00pm

This effort is focused on managing visualization resources which consist of both hardware and software resources. We explore methods to administer a variety of resources in a multi-user environment to allow each user access to dedicated portions of hardware for their work. In addition we discuss ways to manage complex visualization tasks.

back to the agenda



Real Application Performance Analysis on NCSA Systems
Gregory Bauer, Nahil Sobh - NCSA
Wednesday, May 23
10:40am - 11:30am

To reduce turn-around time and memory usage of your application, and to maximize your system's workload, it is essential to understand your application's performance with respect to the system's architecture. We will discuss our experience in analyzing the performance of some real applications using tools that are available to the users of NCSA systems. With performance opportunities in mind, the "best practices" for memory use, communication and I/O will also be discussed.

back to the agenda



NCASSR Cyberinvestigation System
Von Welch - NCSA
Monday, May 21
6:30pm - 8:00pm

Cybercrimes and cyberintrusions are increasingly becoming highly complex events, with perpetrators using multiple systems to obscure their tracks and complex distributed infrastructure constructed from compromised systems to launch attacks. NCSA has led several successful investigations of such incidents and is now building on this experience to develop tools to aid with their resolution. The National Center for Secure Systems Research (NCASSR), led by NCSA, is leveraging our collaboration with the FBI, to develop a system that allows incident responders and law enforcement personnel at distributed sites to work in collaboration to investigate these complex incidents by securely communicating, sharing data and analysis results. We will be demonstrating this NCASSR CyberInvestigation System, which is currently in prototype form.

back to the agenda



 

NCSA Home UIUC Home

National Center for Supercomputing Applications at the
University of Illinois at Urbana-Champaign

All rights reserved.
©2007 Board of Trustees of the University of Illinois.