Long Abstracts of Papers in the AVS '95 Proceedings FYI, these are the long abstracts upon which the short 1-2 sentence presentation descriptions in the Advance Program of AVS '94 were based. -------------------------------------------------------- -----------------DEVELOPERS TRACK ---------------------- --------------------------------------------------------- Chair: Howard Watkins --------------------------------------------------------- Author: Mitchell Roth Arctic Region Supercomputing Center University of Alaska Fairbanks, AK 99775-6020 907-474-5411 (voice) 907-474-5494 (fax) email: roth@arsc.edu Title: AVS for the CRAY T3D Abstract: AVS is a widely used package for scientific visualization which runs on platforms ranging from workstations to Cray Research parallel vector supercomputers. This paper reports on the CRAY T3D parallel implementation and performance of several of the compute intensive AVS modules from the standard libraries. The implementation issues discussed in the paper are: (1) T3D/Y-MP communication, (2) module flow control (3) module generation and (4) parallel computational algorithms for the modules. ----------------------------------------------------------------------- Authors: Karen Woys and Mitchell Roth Arctic Region Supercomputing Center University of Alaska Fairbanks, AK 99775-6020 907-474-6307 (voice) 907-474-5494 (fax) email: woys@arsc.edu and roth@arsc.edu Title: AVS Optimization for CRAY Y-MP Vector Processing Abstract: The AVS source code is modularized and highly portable, but is not vectorized to take advantage of the Cray Research, Inc. (CRI) parallel vector architecture. This project analyzed and optimized the following AVS modules for vector processing on a CRAY Y-MP supercomputer: Read Field, Downsize, Field Math, Field to Mesh, Compute Gradient, and Interpolate. Based on user CPU time, speedups of 10x were generally obtained through vectorization and performance up to 70 MFLOPS (millions of floating point operations per second) was attained in the optimized code. This paper discusses the vectorization and performance measurement techniques that were employed and the ways in which these techniques may be applied to other AVS library modules as well as user written modules. --------------------------------------------------------------- An Application for Visualizing Molecular Dynamics Data Developed under AVS/Express Upul R. Obeysekare, Chas J. Williams, and Robert O. Rosenberg Scientific Visualization Laboratory Center for Computational Science Information Technology Division Naval Research Laboratory Washington, DC 20375-5320 We discuss issues related to developing an application for visualizing data from molecular dynamics simulations. The application is developed using visual programming and user-interface design features of the application development environment of AVS/Express. A suite of AVS5 modules (reported in AVS '94) is being converted to AVS/Express objects that use AVS/Express's new field data scheme. Details related to defining spheres to represent atoms under the new field data scheme is being discussed. Important features for molecular dynamics visualization applications such as picking and highlighting atoms are being tested using the new architecture of AVS/Express. ----------------------------------------------------------------------- Speaker: Walter Schmeing, VISTEC Software GmbH, Berlin, Germany Title: Examining Large Datasets using DBFLD (Field Database) Description: DBFLD has been developed to examine very large datasets stored in compressed form saving a lot of diskspace. Smart readers access these compressed data with the ability of cropping, downsizing, interpolating, mirroring, stretching, slizing, etc. while reading, so you can handle really large datasets now without running out of memory. Interfaces (API's) are available to AVS5-Fields, AVS5-UCD's and AVS/Express-FLD's. ----------------------------------------------------------------------- Dr S. A. Khaddaj Dept. of Computer Science and Electronic Systems, Kingston University, Penrhyn Road, Kingston upon Thames KT1 2EE. email: s.khaddaj@king.ac.uk Atitle An Interactive Design Tool for Scientific Simulations S.A. Khaddaj Department of Computer Science and Electronic Systems, Kingston University, Kingston upon Thames KT1 2EE, United Kingdom Abstract This work is concerned with the development of an interactive graphical environment for scientific simulations. However, as in our previous works [1] [2], special attention is paid to the production, visualisation and animation of the surface morphologies generated from simulation of a material growth technique known as molecular-beam epitaxy (MBE). We are particularly concerned with the use of Application Visualisation System (AVS) as the visual programming interface for the simulations. The simulation/graphics/animation environment is described. New AVS modules created for this specific application are discussed and results are presented. S. A. Khaddaj et al., `The Application of Graphics and Animation Techniques for Large-Scale Simulation', in R. Grebe et al. (eds), Transputer Applications and Systems 93, IOS Press, pp. 244-259, Amsterdam,1993. S. A. Khaddaj et al., `AVS: A Visualisation Environment for Atomic Arrangement and Materials Design', AVS'94, Advanced Visual Systems, Inc., pp. 362-377, Boston, Massachusetts, 1994. ----------------------------------------------------------------------------- AVS/Express Case Study Howard Watkins, John O'Sullivan, David Heath, Peter Yerburgh Intera are in the process of integrating the AVS/Express architecture with their own application framework based on C++ and XVT. Some issues of the integration will be discussed and examples given of the power of the AVS/Express approach. ----------------------------------------------------------------------------- AVS5 and POSTGRES: Large Scale Data Analysis Raymond E. Flanery Jr. Director, Advanced Visualization Research Center Mathematical Sciences Section Oak Ridge National Laboratory phone:(615)574-0630 fax :(615)574-0680 email:flanery@msr.epm.ornl.gov A statistical analysis and browsing application for large data sets has been developed using AVS5 for the user interface and the public domain dbms POSTGRES for the data storage. The Large Scale Data Analysis project requires the use of statistical techniques to extract useful or important information from datasets too large for normal browsing techniques. The information filtered out of these datasets is stored using the POSTGRES dbms. AVS5 is used to build the networks for, and allow interaction with, the statistical filters used to extract the information. It is also used to allow browsing of the resulting data bases. ----------------------------------------------------------------------------- AVS/EXPRESS and PVM: Gas and Oil National Information Infrastructure (GO-NII) Project Raymond E. Flanery Jr. Dr. Bart D. Semeraro Director, Advanced Visualization Research Center Research Staff Mathematical Sciences Section Oak Ridge National Laboratory phone:(615)574-0630 (615)574-3130 fax :(615)574-0680 email:flanery@msr.epm.ornl.gov semeraro@msr.epm.ornl.gov We discuss the use of AVS/EXPRESS as an interface and rendering system for parallelized visualization subroutines, utilizing PVM as the communication protocol. The GO-NII project requires the visualization of seismic data sets which are too large to be manipulated on a single graphics workstation. Typical single shot data sets are ~500MB and a time series of seismic snapshots would consist of hundreds of these data sets. We have implemented a few visualization subroutines with the PVM protocol and interfaced these with AVS. This has allowed us to visualize seismic data sets on a large heterogeneous network of SUN, SGI, DEC and HP workstations. ----------------------------------------------------------------------------- Creating a Scientific Environment with AVS/Express Erin N. Thornton Gary D. Black Tom L. Keller Karen L. Schuchart Chance R. Younkin Donald R. Jones Environmental Molecular Sciences Laboratory Pacific Northwest Laboratory(*) Richland, Washington Productive use of the advanced computational resources available to molecular and environmental scientists requires not only a revolution in computational methods, but also a corresponding revolution in the tools for managing and analyzing computational experiments. The Extensible Computational Chemistry Environment (ECCE') is being developed at Pacific Northwest Laboratory to address these needs. ECCE' is an integrated, comprehensive environment for molecular modeling and simulation. Key components are application systems with graphical user interfaces, chemistry-specific visualization applications, and scientific data management. The graphical user interface assists in the selection of computational parameters, recasting database input queries in terms of scientific requirements, and assisting the user via chemistry-specific help facilities. Visualization tools focus on graphical manipulation and analysis of data from molecular modeling applications. Seamless integration of tools is accomplished through a common architecture for molecular modeling applications that is based on an object-oriented molecular data model. The intent is to allow the execution of complex computational "experiments," provide a framework for extending the computational resources in the systems, and facilitate information sharing. ECCE' is an integration of AVS/Express, an object-oriented database, and an external GUI Builder. AVS/Express serves as the software framework. We discuss the issues of integrating products with AVS/Express and customizing AVS/Express to meet the difficult design requirements of the system being created. The system utilizes an extensively modified AVS network editor. We further discuss the use of a common chemistry data model, and its implementation in AVS/Express. Finally, we address the use of an external GUI builder to create common interfaces to various chemistry applications, and the strategy of integrating these interfaces into the AVS/Express application. (*)The Pacific Northwest Laboratory is a multiprogram national laboratory operated for the U.S. Department of Energy by Battelle Memorial Institute under Contract DE-AC06-76RL0 1830. ----------------------------------------------------------------------------- Developing The Next Generation Nuclear Medical Imaging System With AVS/Express David A. Goughnour Jeffrey A. Hallett, ADAC Laboratories ADAC Laboratories, THE CURRENT WORLD LEADER IN NUCLEAR MEDICAL IMAGING, is currently developing its next-generation diagnostic imaging workstation based on AVS/Express. Nuclear medical imaging is UNIQUE (( a challenging application)) in the imaging world in two respects. From a pure technology viewpoint, nuclear imaging relies heavily on processing of raw image data; using a variety of imaging operations, quantitative results are produced upon which radiologists base diagnoses and treatment procedures for cardiac, oncological, and other types of LIFE-THREATENING illnesses. From the customer standpoint, 90% of the nuclear medicine market is clinically-oriented. These users are generally not highly technical, and are interested only in the aspects of the machine that make it easy to use and allow them to produce the studies they need in the shortest time possible. This paper describes how the various features of AVS/Express are being leveraged by ADAC to not only address the raw computational needs of an advanced nuclear medicine imaging system, but also to produce a simple-to-use, but highly flexible, imaging environment which will appeal to all types of nuclear medicine providers from the base clinician to the advanced researcher. ----------------------------------------------------------------------------- ------------------------------------------------------ ------------------CFD/ENGINEERING TRACK-------------- ------------------------------------------------------ Chairs: Larry Schoof ------------------------------------------------------ Interactive Scientific Exploration of Tokamak Gyro-Landau Fluid Turbulence In A Visual Immersion Environment This work was supported by the USDOE at the Lawrence Livermore National Laboratory under contract number W-7405-ENG-48. G.D. KERBEL, J.L. MILOVICH, and D.E. SHUMAKER National Energy Research Supercomputer Center Lawrence Livermore National Laboratory, Livermore, California 94550, USA gdk@kerbel.nersc.gov R.E. WALTZ General Atomics, San Diego, California 92138-5608, USA, waltz@gav.gat.com} A. VERLO Electronic Visualization Laboratory, University of Illinois at Chicago, Chicago, IL 60607-7053, USA, averlo@eecs.uic.edu} ABSTRACT High performance data parallel algorithms for simulating and visualizing gyro-Landau fluid (GLF) tokamak turbulence in three dimensions have been used in predictive transport scaling studies{WKM} in connection with the Numerical Tokamak Project. The time advancement algorithm is composed of data parallel block structured matrix solvers and multiple instance fast Fourier transforms used in operator splitting and pseudospectral convolution evaluation. The overall performance of the algorithm is optimized by choosing a sequence of data arrangements such that the computation in each arrangement is local and there is a fast data rearrangement algorithm setting the next stage of the sequence. The existence of fast communication kernels connecting certain classes of arrangements helps to determine how to structure the algorithm. The complexity of 3D tokamak turbulence in toroidal geometry requires that advanced visualization techniques be used to understand and communicate results effectively. We have developed an interactive distributed visualization system to examine, compare and display physics results based in part on the same high performance parallel computing and communications algorithms used in the physics simulation. The visualization system provides a wide range of flexibility for image composition and animation, making it easy to select, navigate and frame large data sets for further study and presentation. The system also enables the visual experience of concurrent viewing by collaborators, which is helpful to debug algorithms and suggest meaningful diagnostics. The distributed visualization system currently uses a CM5 as a remote compute and storage server coordinated through CMAVS/AVS{TM} with options for local display ranging from a standard X-server to a 3-D multi-screen stereoscopic projection system using multiple SGI Reality Engine{TM} processors (CAVE{EVL}). In particular, introducing an AVS derived scientific visualization interface into the CAVE environment enables interactive exploration of dimensions of parameter space which might otherwise remain hidden or inaccessible. References {EVL} CAVE is a virtual reality theatre created in the Electronic Visualization Laboratory of the University of Illinois at Chicago. {WKM} R.E. Waltz, G.D. Kerbel and J.L. Milovich, "Toroidal Gyro-Landau Fluid Model Turbulence Simulations In A Nonlinear Ballooning Mode Representation With Radial Modes", Phys. Plasmas 1, 2229 (1994). ------------------------------------------------------------------- NEAR REAL-TIME INTEGRATION OF SUPERCOMPUTING AND RIG TESTS THROUGH HETEROGENEOUS DISTRIBUTED ON-LINE COMPUTATION AND ADVANCED VISUALIZATION SOFTWARE David A. Clark, Aerospace Engineer, Vehicle Propulsion Directorate Army Research Laboratory In most turbine engine testing facilities, the tools and techniques of Computational Fluid Dynamics (CFD) and advanced visualization have never been applied to facilitate (near) real-time analysis of the test hardware. New computer software technology has now been applied which allows Server-To-Client, Remote Procedure Calls (RPC), enabling supercomputers to be called from within the on-line test scanning program. Coupled with advanced visualization software and graphics workstations, it is possible to view the inside of a test while it is being conducted. Such capability can be as valuable to researchers in steering tests as X-rays are to doctors in diagnosing health. Using this on-line system, a full turbomachine (compressor) has been visually analyzed by interpolating pressure and temperature instrument rakes to give a full flow-field view of the engine (compressor). All data values at each grid cross-section are non dimensionalized at each grid cross section and viewed at varying ranges of iso-distortion surfaces. Regions of low or high energy can be seen as they proceed through the compressor stages. A full range of capabilities are displayed for both temperature and pressure using computer animation techniques recorded to video. Such views are unique and may provide extra information to help understand inlet distortion as it relates to stall margin. If better stall management can be attained, compressors can be operated safer and possibly be allowed to operate closer to "peak efficiency". More efficient engines have a direct impact on operating costs and promise tangible monetary savings. Secondly, CFD efforts are described in conjunction with the use of RPC to supercomputers. A velocity gradient meridional, quasi-3D solution to five blade rows is shown in context with rotating blade rows and shown on video. The value of the computer work is all generic and can be applied in almost any scientific area where on-line computer systems are used. ------------------------------------------------------ Spectral Element Modeling of Seismic Wave Propagation G. Padoan, A. Pregarz, and E. Priolo Osscivatorio Geofisico Sparimentalc tel: 39 40 2140260 padoan@gem755.ogs.trieste.it Abstract: The numerical simulation of the propagation of both acoustic and elastic wave fields in hetereogeneous media is presented by an animated visualization. The wave fields are computed numerically by the spectral element method, according to a high-order finite element approach. The elements of the unstructured grid contain many edge and internal nodes, depending on the order of the polynomial base; for instance, for a six order polynomial each element has 45 nodes, of which 20 are edge nodes and 25 internal nodes. In order to manage unstructured data, AVS has some pre-defined low-order cells. One way to represent by AVS the wave fields generated by high-order methods based on unstructured grids is to define new high-order cells, and to implement well suited graphics modules. Advance: This paper shows the results of a simulation of the propagation of both acoustic and elastic wave fields in a hetereogeneous media, where the unstructured high-order grid data is handled by defining new high-order cells, and then implementing well-suited graphics modules for them. ------------------------------------------------------ ------------------------------------------------------ --------------------RESEARCH TRACK-------------------- ------------------------------------------------------ Chair: Chuck Hansen ------------------------------------------------------ Real-Time Visual Control of Numerical Simulation Upul Obeysekare, Fernando Grinstein, Gopal Patnaik, and Chas Williams Laboratory for Computational Physics and Fluid Dynamics Naval Research Laboratory Washington, DC 20375-5320 We address relevant issues and difficulties involved in the practical implementation of real-time visualization with emphasis on interactive control of numerical simulations. Important issues governing the implementation of this technique such as network data transfer speeds, asynchronous module execution, architecture neutral implementation, visual programming environment across heterogeneous computer architectures, and user-interface design for visual control of the simulation are being addressed. Strategies to overcome difficulties associated with the implementation of the concept are analyzed in the context of selected numerical simulations implemented under AVS's heterogeneous computing environment using remote Cray Supercomputers and local graphics workstations. Future requirements for collaborative visualization in this environment is also being addressed. ------------------------------------------------------------------- "Visualization and feature extraction in isotropic Navier-Stokes turbulence" Victor M. Fernandez and Norman J. Zabusky Department of Mechanical and Aerospace Engineering and CAIP Center, Rutgers University, Piscataway, NJ 08855} Smitha Bhat and Deborah Silver Department of Electrical and Computer Engineering and CAIP Center, Rutgers University, Piscataway, NJ 08855 Shi-Yi Chen IBM T.J. Watson Research Center and Theoretical Division and Center for Nonlinear Studies, Los Alamos National Laboratory, Los Alamos, NM 87545 ADVANCE: Feature extraction and data reduction algorithms enhance physical understanding in visualization of Navier-Stokes turbulence and also provide an efficient way to deal with large datasets. ABSTRACT: We present feature extraction and data reduction algorithms and provide an insight in the types of problems that arise in dealing with large datasets, obtained in Navier-Stokes turbulence simulations with a 512x512x512 mesh resolution. The developed tools are based on thresholding, object segmentation and low order ellipsoidal quantifications and are applied to the search of coherent vortex structures associated with maxima events in the turbulence field. We underline the importance of sharing tasks between the supercomputer (CM5) and the workstation (SGI Onyx), where each machine may work more efficiently at different stages of the data processing. We obtain visualizations that show the structure of the dominant coherent objects. The reduced representations employed make it possible to examine different types of fields for possible correlations. The quantification of the objects identified by the feature extraction algorithms, should contribute to the building of models that consider both coherent structures and the random background observed in Navier-Stokes turbulence.} ------------------------------------------------------------------- Wes Bethel LBL Modular Virtual Reality Visualization Tools While the debate continues over whether or not the term "Virtual Reality" is an oxymoron or pleonasm, those of us tasked with developing or using visualization software more often than not must eschew this largely philosophical debate and focus our attention on more practical matters: usable software and scientific progress. VR implementations range from fully immersive to desktop systems. The goal of each of these implementations is to provide a user interface in which a human can interact with a computer model in a way which is intuitive, "easy to use", interesting and engaging. We describe an approach to desktop VR geared towards the user of modular scientific visualization systems which combine inexpensive yet practical VR input devices with a methodology for using the data generated by these devices in a variety of ways. For example, a tracker could be used in one context to position a viewpoint, but in another, to orient a slice plane. Most VR devices generate information which is six-dimensional: position and orientation. Along with that information, many have one or more buttons which can be used to generate boolean events. The methodology presented shows one way to manage this type of input, and capitalizes on the strengths of modular visualization environments by allowing the user freedom of choice about the way in which the VR input device data is used in the visualization network. ------------------------------------------------------------------- Quantitative Analysis of Reconstructed Rodent Embryos Andy R. Haas, Richard A. Strelitz, Ph.D., William S. Branham, Daniel M. Sheehan, Ph.D. System Analyst, R.O.W. Sciences, Inc. Division of Reproductive and Developmental Toxicology National Center for Toxicological Research Jefferson, Arkansas, U.S.A., 72079 ahaas@fdant.nctr.fda.gov (501) 543-7011 Advance Detailed insight into the development of rodent embryos is being achieved through quantitative volume analysis at National Center for Toxicological Research. Abstract The Application Visualization System (AVS) provides an interactive rendering environment suitable to assess size (volume) changes in fetal biological systems. Altered fetal growth is likely to result in birth defects in offspring. However, early effects on growth are difficult to measure because of the small size of fetal organs. Using laser scanning confocal microscopy, embryos can be "optically sectioned" to reveal organ structure. Analysis of these optical sections, either singly or as reconstructed embryo objects, can detect early abnormal embryo growth. AVS modules have been developed which model and annotate organ objects, support organ selection, and measure volumes. Selecting an organ displays both an annotation describing the organ and the organ volume. General purpose volume analysis is also obtained through a compute volume module. Volume and surface area measurements may be obtained from any AVS surface geometry (e.g., isosurface output). A slice module has also been developed for researchers to investigate complexities in an organ's surface structure such as folds or deep concavities. -------------------------------------------------------------------------- Real Time MPP 3-D Volumetric Visualization: Medical Imaging on the Cray T3D with AVS Wolfgang Kraske Northrop Grumman Corp. Chris Asano Cray Research Inc. Abstract A joint research effort of Northrop Grumman Corp. (NGC) and Cray Research Inc. has produced the first real time implementation of the Advanced Visual System (AVS) volumetric ray tracing algorithm, tracer, on a Cray T3D computer. The AVS ray tracer was further enhanced for clinical diagnositic therapeutic applications with unique NGC morphological tissue classification algorithms originally developed for military target acquistion applications. This applications demonstrates a sustained .25 second renderings of full color with transparency on 256**3 biomedical X-ray computed tomography and magnetic resonance data sets. Advance We have integrated state of the art military target acquistion technology with AVS 3D visualization tools on a hetrogeneous computing environment to achieve a real-time implementation of enhanced visualization and medical tissue classification on the Cray T3D MPP. ----------------------------------------------------------------------------- -------------------------------------------------- --------ENVIRONMENTAL/EARTH SCIENCES TRACK-------- -------------------------------------------------- Chairs: Theresa Rhyne & Wes Bethel -------------------------------------------------- AVS in Climate Research Dr. Joachim Biercamp Head of Visualization Group Deutsches Klimarechenzentrum GmbH (DKRZ) (German Climate Computing Center) Bundesstr. 55 D-20146 Hamburg Germany The German Climate Computing Center (DKRZ) provides the computer power for quantitative computation of complex climate processes with sophisticated, realistic models. DKRZ also tries to provide the scientists with easy access to visualization tools, helping them to improved insight into the huge amount of data resulting from these computations. The paper will show AVS5-visualizations of oceanographical, metorological and geophysical data. Applications include the global warming, the global carbon cycle and El Nino. We will also try to present first results obtained with AVS6. ----------------------------------------------------------------------- CLIMATE SIMULATION STUDY III: Supercomputing and Data Visualization Philip C. Chen Fujitsu America, Inc. 3055 Orchard Drive San Jose, California 95134 U.S.A. email: pchen@fai.com During the last two years, the Community Climate Model (CCM), a numerical model developed and updated by the National Center for Atmospheric Research (NCAR) scientists, has been ported to the Fujitsu VPX240 supercomputer. Data generated by 3 versions of CCM were used by exploring data visualization techniques suitable for climatological studies, and these techniques were reported in the AVS '93 and '94 conferences. Current research efforts involve using a supercomputing-visualization facility for climate simulation and data visualization applications. The facility includes: a Fujitsu vector supercomputer VPX240 and a vector-parallel-processing supercomputer VPP500, workstations, printers, video recorders, a film recorder, and a scanner. The supercomputers, workstations, as well as video and graphics input/output devices are connected by networking systems. The supercomputers have been used mainly for data generation, and workstations which are connected to supercomputers have been used to visualize data. The current version of CCM: CCM2 has recently been ported to the Fujitsu VPP500 supercomputer. The VPP500 is a highly parallel, distributed memory supercomputer, and it includes multiple processing elements (PE's) with each PE having 1.6 GFLOPS performance. With many PE's, the VPP500 has a proven record of achieving more than 100 GFLOPS. It is expected that the CCM2 will run better on VPP500. A possibility of doing data visualization while generating data with VPP500 is being investigated. AVS macro-module networks were developed on a SGI workstation to study basic climatological parameters and derived parameters. The slowness of AVS has been noted especially when time-animation and volume-visualization are involved. AVS compute-intensive remote modules have been ported to the Fujitsu VPX-series computer at the University of Manchester, and they have been used successfully with AVS modules installed in workstations. In the future, these AVS remote modules will be made available to the on-side supercomputer. At that time, AVS execution speeds of the supercomputer and the workstation will be compared. ----------------------------------------------------------------------- Methods of Constructing a 3D Geological Model from Scatter Data Jennifer Horsman Earth Sciences Division Lawrence Berkeley Laboratory 1 Cyclotron Road Berkeley, CA 94720 jlhorsman@lbl.gov Wes Bethel Information and Computing Sciences Division Lawrence Berkeley Laboratory 1 Cyclotron Road Berkeley, CA 94720 ewbethel@lbl.gov Abstract Most geoscience applications, such as assessment of an oil reservoir or hazardous waste site, require geological characterization of the site. Geological characterization involves analysis of spatial distributions of lithology, porosity, etc.. Geoscientists often rely on two-dimensional visualizations for analyzing geological data. Because of the complexity of the spatial relationships, however, we find that a three-dimensional model of geology is better suited for integration of many different types of data and provides a better representation of a site than a two-dimensional one. Being able to easily manipulate a large amount of heterogeneous data increases the level of interactivity by providing the geoscientist with the opportunity to detect and visually analyze spatial correlations and correlations between different types of data, and thus leads to an increased understanding of the data. A three-dimensional model of geology is constructed from sample data obtained from field measurements, which is usually scattered. To create a volume model from scattered data, interpolation between points is required. The interpolation can be computed using one of several computational algorithms. Alternatively, a manual method may be employed, in which an interactive graphics device is used to input by hand the information that lies between the data points. For example, a mouse can be used to draw lines connecting data points with equal values. The combination of these two methods presents yet another approach. In this study, we will compare selected methods of three-dimensional geological modelling. We used a flow-based visualization system (AVS) to construct the geological models computationally. Within this system, we used two modules, scat_3d and scatter_to_ucd, to interpolate scattered data. These modules will be compared with the combined manual and computational approach. To demonstrate this method, we used a geological modelling system. ----------------------------------------------------------------------- ------------------------------------------------------------- --------------------------IMAGING TRACK---------------------- ------------------------------------------------------------- Chair: Todd Rodgers ------------------------------------------------------------- Maximum Detection Range of Target Edge as a Function of Variable Precipitation and Cultural Obscurants Clifford A. Paiva, MS Research Branch, Countermeasures Division Warfare Systems Department Naval Surface Warfare Center Dahlgren Division (703) 663-4781 Problem Performance of new generation precision guided munitions (PGM) and smart weapons is often unpredictable and unreliable for the global range of battlefield environments which include natural and hostile cultural (man-made) countermeasures. Field test results do not extrapolate well to the full range of operational conditions.(1) The primary inhibitory factor for resolution of automatic target recognition (ATR) performance problems has been the inability to quantitatively characterize target discrimination algorithms which include backgrounds and countermeasures (TDB/C). This study is directed to addressing one of the more serious problems in ATR algorithm performance: efficient segmentation of imagery containing high standard deviation (clutter noise) values, generated by variable precipitation, turbulence and smoke. Approach Four precipitation rates (75,50,25, and 5 mm/hr) through turbulence, humidity and obscurants were selected for a perturbation analysis, from which available target edge intensities, and first and second order statistics were obtained. A high differential scattering cross section to total cross section scattering from the segmented scene is assumed. This effectively reduces the grayscale intensity of the target (M60 Main Battle Tank) and allows an assessment of target detection via morphological image processing techniques. The final edge intensities and statistics were then available for automatic target recognition (ATR) geometric pattern referencing, as well as correlation mapping routines. Maximum detection sensor-to-target range for variable rain rates, turbulence and obscurants, were summarized. Image processing techniques for this analysis included erosion/dilation and open/close operations, as well as region-growing image segmentation. Four LM8 smoke grenades generated obscurants which moved normal to the line-of-sight (LOS), and through the turbulence. The imagery is real in 9-12 micron bandpass taken at the US Army Keweenaw Research Center. Sensor-to-target range commences at 160 meters (maximum) and closes to 600 meters (minimum). Pixels on the partially obscured extracted edges, are then counted and plotted versus range and precipitation constants. Statistics (stand deviation and kurtosis) are plotted. Results The results indicated that the number of pixels-on-edge (edgels) vary as a function of changes in three independent variables: (1) range; (2) precipitation rate; and (3) position of obscurant (smoke) to the target. Decreased target edge mean and increased standard deviation, as a function of increased precipitation, resulted in reduced detection and classification potential for automatic target recognition (ATR) algorithms. Although some pattern referencing information is present for Marr-Hildreth type zero-crossing edge detectors, segmentation operations were severely stressed, and attempts to reduce high and low clutter frequencies were not successful beyond 1000 meters sensor-to-target, particularly for the partially obscured target. Nevertheless, although the target intensity was significantly less than the smoke intensity, due to the high target intensity gradients and low smoke gradients, the final segmented, edged-enhanced scene, successfully revealed only the target (M60 MBT) in ranges less than 1000 meters. Smoke obscurant was effectively filtered from imagery as the missile closed on the target. Actual count of edgels (versus precipitation rate) indicated strong dependence on obscurant cloud position relative to target, precipitation rate, and range. An eight minute video of the scenario is provided. The results gives some insight and perspective regarding maximum detection ranges which may be expected in severe natural and cultural cluttered environments. ------------------------------------------------------------- A GENERALIZED CONVOLVER MODULE Johan Wiklund and Hans Knutsson Department of Electrical Engineering Linkoeping University S-581 83 Linkoeping SWEDEN A procedure to perform convolutions on multi-dimensional data with arbitrary filter kernels is a basic tool in image and signal processing. Typical input data are 1D signals, 2D images, 3D volumes, 3D spatio-temporal image sequences and 4D volume sequences. Each coordinate, (pixel, voxel, toxel), can contain a scalar value or a vector. If the kernel and/or the input data has a vector length larger than one, a generalized convolution is needed. In this case the multiplication in the convolution is changed to a vector combination. Examples of vector valued input data are color (RGB), 2D vector fields, 3D vector fields, tensor fields etc. A scheme for performing generalized convolutions is presented. A flexible convolver, which runs on standard workstations, has been implemented. It is designed for maximum throughput and flexibility. The implementation incorporates spatio-temporal convolutions with configurable vector combinations. It can handle general multilinear operations, i.e. tensor operations on multidimensional data of any order. The input data and the kernel coefficients can be of arbitrary vector length. The kernel coefficients can be scattered, i.e they don't need to be uniformly placed inside a box. The computational cost increases lineary with the number of kernel coefficients, it does not depend on the size of the kernel bounding box. A region of interest (ROI), e.g. a spatial rectangle in the input over which the convolution should be applied, can be defined. Subsampling is user selectable and decreases the computational cost for the convolution. There are two basic classes of filters, FIR (finite impulse response) and IIR (infinite impulse response). Both types of filters are supported by the convolver, it is configurable for IIR filters in the time dimension. The implementation is done as a C-library and a graphical user interface in AVS (Application Visualization System). ----------------------------------------------------------------------- Marquess Lewis (melewis@tasc.com) TASC 55 Walkers Brook Drive Reading, MA 01867 (617)942-2000 ABSTRACT Radiometric and Geometric Adjustment of Airborne Spectrometer Data within the AVS Environment Marquess E. Lewis MTS, Signal and Image Technology Division TASC, 55 Walkers Brook Drive, Reading MA 01867 Multispectral and hyperspectral imagery collected from airborne platforms is finding increasing use in mineralogical, agricultural, and oceanic studies. Prior to application specific analyses, these data must be corrected for geometric distortions induced by the instrument optics and collection geometry and the radiometric response characteristics of the sensor. TASC has recently fielded a ground station for the processing and analysis of airborne spectrometer data. A component of this ground station allows scientists to interactively perform "what if" kinds of analyses using the AVS environment with a large set of custom modules. With in the AVS environment radiometrically and geometrically adjusted data may also be produced. This paper describes the overall TASC MIDAS mission and then focuses on the data normalization process as implemented within AVS. Results using recently collected imagery will be shown. ----------------------------------------------------------------------- TITLE: Image Processing in the Spatial and Frequency Domains Stephen L. Schultz System Programmer, Center for Imaging Science Rochester Institute of Technology One Lomb Memorial Drive Rochester, New York 14623 (716) 475-5294, slspci@rit.edu ABSTRACT: The Center for Imaging Science uses AVS to bridge the gap between the numerous imaging and remote sensing packages utilized at the Center. In addition, the Center's library of AVS modules provides additional functionality missing in the various packages, the most notable of which is frequency domain image processing. AVS allows the researchers at the Center to examine the feasibility of applying various spatial or frequency domain solutions to image processing problems with a minimal amount of time or effort. The Center's module library includes support for most of the image storage formats used by the various packages in use. This presentation will demonstrate some of this functionality along with the considerations used in developing the module library. Concrete examples from research projects conducted by the Center will be included. ADVANCE: A demonstration of the AVS modules developed at the RIT Center for Imaging Science used to bridge the gap between the numerous imaging and remote sensing packages utilized at the Center along with research project examples and considerations taken in developing the modules. ----------------------------------------------------------------------- -------------------------------------------------------- --------------COMMERCIAL TRACK-------------------------- -------------------------------------------------------- Chair: Graham Walker -------------------------------------------------------- CHALLENGES IN INFORMATION VISUALISATION Dr Graham Walker graham.walker@bt-sys.bt.co.uk Advanced Applications and Technologies BT Laboratories Martlesham Heath Ipswich IP5 7RE UK We are surrounded by an ever-growing, ever-chaning world of data. However, the value of this data is not intrinsic, but lies in enabling us to make more informed decisions and in increasing our shared knowledge and understanding. In this paper, we explore the critical role for visualisation in bridging the gap between the abstract analytical world of data and the digital computer, and the real, analogue world of human problems and experience. Our discussion is structured around four Challenges, derived from current trends in data and business practice: a growing volume of data with declining information content, more complex data analysis tools and models; increasingly abstract data tools and models; and a wider, less specialised audience. We illustrate our comments with examples from worl on visualisation of telecommunications data at BT Laboratories. The majority of these examples have been developed using AVS. ----------------------------------------------------------------------- Using AVS as a virtualization tool for WWW Jeff Wang MCNC/North Carolina Supercomputing Center 3021 Cornwallis Road Research Triangle Park, NC 27709 jfwang@robin.ncsc.org The World Wide Web (WWW) and internet information technology have added a new dimension to the application area of AVS. The lack of visualization capability of the WWW tools can be compensated by using AVS. The images generated by AVS can be browsed and archived by WWW. This paper presents a prototype of using NCSA Mosaic to run AVS for demostration purpose. AVS can generate both GIF format image and MPEG data file, NCSA Mosaic and the multimedia viewers can access these image files and produce visualization results in nice quality. The new version of NCSA Mosaic has provided a common client interface (CCI) that allows more interactive AVS image browsing, and it opens up new potential application areas such as image on demand, online tutorial, and collaboratory in AVS development. The paper also evaluates the efficiency and pitfall of the connection between AVS and WWW. ----------------------------------------------------------------------- An AVS Interface to the Aurora(TM) Dataserver Charles Falkenberg Department of Computer Science University of Maryland College Park, MD Mike Achenbach Xidak Corporation Palo Alto, CA Ravi Kulkarni Advanced Visualization Laboratory University of Maryland College Park, MD Vince Patrick Prince William Sound Science Center Cordova, AK Abstract: In this paper we present a framework for integrating AVS with a scientific database management system consisting of the Aurora Dataserver from Xidak. The Aurora Dataserver includes a wide range of database functionality for spatially coordinated data. This functionality includes SQL queries, a high level data model(ie. netCDF), historical meta-data, and transaction management in a client server architecture. Our long term objective is to use AVS and the Aurora Dataserver to form an integrated view of the Prince William Sound (PWS) ecosystem in Alaska. The datasets used with the Aurora Dataserver are taken from a diverse collection of oceanographic and biological data from the PWS Ecosystem Assessment project (SEA) undertaken as a result of the Exxon Valdez oil spill. We envision using the Aurora Dataserver for temporal and spatial queries and AVS networks to help test different hypothesis, models, and compare different datasets. Our initial efforts have been the development of two modules which provide read/write mapping from AVS fields to Aurora datasets as well as allowing SQL queries, sub-sampling, and region extraction. We will describe the functionality of the Aurora Dataserver and its underlying data model, the interface to our AVS modules, and the longer term enhancements necessary for support of the SEA project. ----------------------------------------------------------------------- Wayne Haidle Montana-Dakota Utilities Co. Correlating Time-Based Power System Parameters via Preprocessing UNIGRAPH Command File Templates The point and click capability of UNIGRAPH is adept at producing a command file to replicate a given interactive process. However, the command file is not flexible in adapting to a variety of batch process interrogations. A command file preprocessor can be utilized to overcome this limitation via creating a unique command file from the command file template. ----------------------------------------------------------------------------- ------------------------------------------------------------ -------------------OIL & GAS TRACK-------------------------- ------------------------------------------------------------ Chairs: Mike Ray & Annette Walsh ------------------------------------------------------------ Title: Visualization of Multi-component Saturation Distributions in Oil Reservoir Models Authors: Dong Ju Choi and Mitchell Roth Arctic Region Supercomputing Center University of Alaska Fairbanks, AK 99775-6020 907-474-6307 fax: 907-474-5494 email: choi@arsc.edu and roth@arsc.edu Abstract: Sophisticated oil reservoir models simulate pressures and saturations for dozens of hydrocarbon components in a 3D geometry. Visualizing the composition of the components is difficult using the standard AVS color table when more than two components are present. For multi-component models a transformation of the standard color table is necessary. One approach is to use a color table built from a regular polygon, where each vertex represents a pure component. In the current study, a triangular color table is employed to display oil, gas, and water saturations resulting from the 3D time dependent simulation of an oil field employing gas injection for enhanced recovery. This paper will show the process of converting the triangular color table into an AVS color table and its application to the visualization of oil reservoir models. ----------------------------------------------------------------------- ABSTRACT Use of AVS in the Uncertainty Analysis of the Depth Structure of an Idrocarbon Reservoir Paolo Ruffo, Livia Bazzana, Ernesto Della Rossa, Rita Colombo Senior Professional, Integrated Interpretative Applications Dept. AGIP SpA Email: ruffo@agip.geis.com In the field of the depth conversion of seismic time maps, AGIP has developed a methodology that allows the evaluation of the uncertainty of the depth model. This methodology was translated into a software package (called GEODE) that we created using AVS as a development tool. This allowed us to obtain several advantages, in fact the final package is: user friendly, easy to maintain, integrated in our processing flow and enough flexible to be fitted to variable requirements and to future modifications. The conventional depth conversion technique produces a unique depth map starting from a seismic time map and a seismic velocity map. In the GEODE approach the conventional unique depth map is interpreted as an average result, that is: its difference with the unknown reality is in average zero. The basic principle applied in GEODE is to start from the knowledge that the seismic velocities are uncertain and to apply a geostatistical technique to evaluate the effect of the velocity uncertainty on the depth uncertainty. In a way similar to that of the classical montecarlo technique, we produced a number of geostatistically simulated velocity fields, all compatible with all input velocity data, their uncertainty and their spatial correlation. Using all these simulated velocity field in the depth conversion process we produced several simulated depth models, instead of just one, as with a conventional approach. The final analysis and quantification of the uncertainty of the depth model is based on an AVS interactive graphical interface, so that the user may easily synthesise and understand the characteristics of depth stochastic model. Obviously one of the result of this approach is that we can produce a distribution of volumes (instead that just one number) and use it in the process of economic evaluation of the Reservoir. ADVANCE The use of AVS modularity improved the realisation of a package that allows to estimate and analyse interactively the uncertainty of the depth model and of the volume of a Potential Reservoir. ----------------------------------------------------------------------- Title: Developing an Oil Reservoir Simulator Post-processor Using agX/Toolmaster & UIM/X Speaker: Dr. David Pottinger, Principal Software Engineer, Consultancy Services, AEA Technology Dorchester Dorset, UK DT2 8DH email: david.pottinger@aea.orgn.uk tel: (+44) 1305 202896 fax: (+44)1305 202110 Abstract: An application has been developed by AEA Technology Consultancy Services to fulfill a requirement to visualize the results of reservoir simulations. The emphasis has been on the day-to-day requirements of engineers, namely to provide line chart, contour and cross-section displays. Whilst many current applications provide high-end three dimensional visualization, a lack of tools able to support such displays has been identified. agX/Toolmaster was chosen as the most appropriate graphics library due to its range of functionality, speed, quality of displays and ease of use, as well as the range of hardcopy devices supported. The application was developed in two distinct parts - the graphical user interface, using UIM/X, and the functionality, using agX/Toolmaster. This allowed the interface and the functionality to be designed, implemented and tested independently in parts. Communication between the two parts was achieved using a command language - = the interface generated commands which were passed to a command handler and processed accordingly. This approach allows the application to run both interactively, driven by the interface, and in background, driven by a command file. The use of Motif and Toolmaster ensures that the application is easily ported to a range of platforms. Using UIM/X allowed the development of prototype interfaces which users commented on prior to final implementation. The range of options provided by Toolmaster allowed the images to be easily customized through a range of dialog options. The design of the dialogs has been aimed at providing providing sensible initial default settings, ease of use. Extensive use of resources allows both the interface and the command language to be customized for overseas markets. Advance: A presentation of techVISION - an application developed using agX/Toolmaster for visualisation of oil reservoir simulation results, and UIM/X to provide a powerful and easy to use graphical user interface. ----------------------------------------------------------------------- ------------------------------------------------------------ -----------------------MEDICAL TRACK------------------------ ------------------------------------------------------------ Chair: Marc Kessler ------------------------------------------------------------ Loyd Myers Dept. of Biological Structure, University of Washington (206) 543-5480 myers@biostr.washington.edu VOLUME RENDERING OF MULTI-SPECTRAL MR DATA FOR BRAIN LANGUAGE SITE MAPPING Loyd M. Myers, Jeff Prothero, James F. Brinkley We describe a technique for visualizing cortical anatomy in relationship to arteries and veins in order to map language sites in a 3D coordinate system. The technique requires three sets of MRI data, optimized for the three tissue types, to be acquired pre-operatively, registered, combined, and finally rendered in a single image. The three datasets are registered from machine coordinate information contained in the MRI headers. Semi-automated segmentation, using an adaptive region growing algorithm followed by morphological dilation, is used to produce a rendering mask. The three data sets are then masked, combined, and rendered to produce the final image. The technique was applied to three patients, and the rendered results compared to intraoperative photographs of the cortical surface. In each case, the corrolation of landmarks between the photographs and the rendered images was sufficient to allow a neurosurgeon to correctly locate the language sites in the rendered image. This technique is impemented by a combination of native AVS modules, AVS-based user applications, and in-house code which is accessed as a server through AVS client modules. This approach has allowed us to develop a complex heterogenous application from pieces written by programmers working independantly in different programming environments. ADVANCE: A technique for visualizing cortical anatomy in relationship to arteries and veins from MR data in order to map language sites in the human brain is described. ----------------------------------------------------------------------------- Ted Beatie Mass General Center for Imaging and Pharm Research Charleston, MA 617-726-7834 email tcb@cipr.mgh.harvard.edu (title only for now) Comparison of 2D and 3D segmentation techniques for volume determination in CT phantoms ----------------------------------------------------------------------------- An AVS-based System for Optimization of Conformal Radiotherapy Treatment Plans J.J. Kim, M.L. Kessler, N. Dogan, and D.L. McShan The University of Michigan, Ann Arbor, MI 48109 While it is now possible using computer-controlled treatment machines to employ larger numbers of fields to better conform dose to a target volume, current interactive planning tools are not effective for handling the increased degrees of freedom. Therefore, some level of computer-aided optimization is needed. Towards this goal, we have developed an AVS-based system which allows the treatment planner to rapidly specify a range of possible beam orientations and weights and to interactively construct cost functions to be optimized and constraints to be satisfied. Examples of optimization criteria include biophysically motivated cost functions using normal tissue complication and tumor control probabilities. Once the necessary information is given, an optimization is begun and intermediate results, such as cost history, beam weights, and dose-volume histograms are displayed. At any time during an optimization, the process can be interrupted, any parameter or criteria modified,and then restarted. By considering only dose points within the tissues under consideration, it takes only a few minutes to perform thousands of iterations of beam weightings and cost evaluations. The blend of automated and interactive optimization allows the treatment planner to manage the large number or degrees of freedom possible. This presentation will describe the details of this system and provide examples of its use. This work was supported in part by NIH grant PO1-CA59827. ----------------------------------------------------------------------------- Chris Siegel NYU School of Medicine 212-263-5744 Creating 3D Models from Medical Data using AVS. Because of its easily extended set of processing modules, AVS has grown to be a very powerful tool for the extraction of object models from medical images. However, not all of the challenges in the task of accurately isolating the objects of interest without human intervention have been met. I will explain what I have been able to accomplish using AVS as a processor for CT and MRI patient data, and what stumbling blocks and brick walls I have come up against. -----------------------------------------------------------------------------