Nickolas L. Faust

Real-time, continuous level of detail rendering of height fields

We present an algorithm for real-time level of detail reduction and display of high-complexity polygonal surface data. The algorithm uses a compact and efficient regular grid representation, and employs a variable screen-space threshold to bound the maximum error of the projected image. A coarse level of simplification is performed to select discrete levels of detail for blocks of the surface mesh, followed by further simplification through repolygonalization in which individual mesh vertices are considered for removal. These steps compute and generate the appropriate level of detail dynamically in real-time, minimizing the number of rendered polygons and allowing for smooth changes in resolution across areas of the surface. The algorithm has been implemented for approximating and rendering digital terrain models and other height fields, and consistently performs at interactive frame rates with high image quality.

View-dependent multiresolution splatting of non-uniform data

This paper develops an approach for the splat-based visualization of large scale, non-uniform data. A hierarchical structure is generated that permits detailed treatment at the leaf nodes of the non-uniform distribution. A set of levels of detail (LODs) are generated based on the levels of the hierarchy. These yield two metrics, one in terms of the spatial extent of the bounding box containing the splat and one in terms of the variation of the scalar field over this box. The former yields a view-dependent choice of LODs while the latter yields a view-independent LOD based on the field variation. To show the utility of this general approach it is applied to a set of application data for a whole earth environment and some test data. Performance results are given.

Acquisition and display of real-time atmospheric data on terrain

This paper investigates the integrated acquisition, organization, and display of data from disparate sources, including the display of data acquired in real-time. In this case real-time acquisition and display refers to the capture and visualization of data as they are being produced. The particular application investigated is 3D dynamic atmospheric data on terrain, but key elements presented here are applicable more generally to other types of real-time data. 3D Doppler radar data are acquired and visualized with global, high resolution terrain. This is the first time such data have been displayed together in a real-time environment and provides the potential for new vistas in forecasting and analysis. Associated data such as buildings and maps are displayed along with the weather data and the terrain. A global hierarchical structure makes these disparate data available for integrated visualization in real-time. Requirements for effective 3D visualization for decision-making are identified, and it is shown that the applications presented meet most of these requirements.

Situational visualization

In this paper, we introduce a new style of visualization called Situational Visualization, in which the user of a robust, mobile visualization system uses mobile computing resources to enhance the experience and understanding of the surrounding world. Additionally, a Situational Visualization system allows the user to add to the visualization and any underlying simulation by inputting the users observations of the phenomena of interest, thus improving the quality of visualization for the user and for any other users that may be connected to the same database. Situational Visualization allows many users to collaborate on a common set of data with real-time acquisition and insertion of data. In this paper, we present a Situational Visualization system we are developing called Mobile VGIS, and present two sample applications of Situational Visualization.

Semiautomated and interactive construction of 3D urban terrains

We have developed a set of tools that attack the problem of rapid construction of 3D urban terrains containing buildings, roads, trees, and other features. Heretofore, the process of creating such databases has been painstaking, with no integrated set of tools to model individual buildings, apply textures, place objects accurately with respect to other objects, and insert them into a database structure appropriate for real-time display. Since fully automated techniques for routinely building 3D urban environments using machine vision have not yet been entirely successful, our approach has been to build a set of semiautomated tools that support and make efficient a human interpreter, running a PC under Windows NT.

Building the visual Earth

Over the past several years there has been a broad effort towards realizing the Digital Earth, which involves the digitization of all earth-related data and the organization of these data into common repositories for wide access. Recently the idea has been proposed to go beyond these first steps and produce a Visual Earth, where a main goal is a comprehensive visual query and data exploration system. Such a system could significantly widen access to Digital Earth data and improve its use. It could provide a common framework and a common picture for the disparate types of data available now and contemplated in the future. In particular mcuh future data will stream in continuously from a variety of ubiquitous, online sensors, such as weather sensors, traffic sensors, pollution gauges, and many others. The Visual Earth will be especially suited to the organization and display of these dynamic data. This paper lays the foundation and discusses first efforts towards building the Visual Earth. It shows that the goal of interactive visualization requires consideration of the whole process including data organization, query, preparation for rendering, and display. Indeed, visual query offers a set of guiding principles for the integrated organization, retrieval, and presentation of all types of geospatial data. These include terrain elevation and imagery data, buildings and urban models, maps and geographic information, geologic features, land cover and vegetation, dynamic atmospheric phenomena, and other types of data.

High-fidelity infrared scene simulation at Georgia Tech

The Georgia Tech Research Institute has for more than fifteen years developed and used digital scene models for IR simulation applications. Initially focusing on synthetic scenes of small extent but very high resolution (less than one meter), more recently emphasis has shifted to larger scenes derived from measured data sources with resolution at one meter or slightly greater. One reason for the shift in emphasis has been the emergence of the GTSIMS simulation environment, in which digital IR seeker and missile models and models of other EO/IR sensor systems used in tactical missile engagement scenarios require larger scene extents (typically three to ten kilometers on a side) because of their potential viewing geometries and fields of view. In GTSIMS these sensor and missile models are integrated in a unified software system with the IR scene models and the image rendering software that has been developed along with them. The GTSIMS missile engagement capabilities, including many aspects of scene configuration and signature prediction, are tied together through a graphical user interface called XGTSIMS. This paper will discuss recent IR scene models developed for GTSIMS, from the methodologies used to create the data sets behind the models to the use of these models in GTSIMS via XGTSIMS, then will proceed to discuss current and planned efforts toward real-time image generation of large, complex scenes for IR simulation purposes.

4D symbology for sensing and simulation

The Armys Common Picture of the Battlefield will produce immense amounts of data associated with tactical goals and options, dynamic operations, unit and troop movement, and general battlefield information. These data will come form sensors (in real-time) and from simulations and must be positioned accurately on high-fidelity 3-D terrain. This paper is associated with the Armys 2-D symbols for operations and tactics so that the information content of this symbolic structure is retained. A hierarchy is developed based on military organization to display this symbology. Using this hierarchy, even complex battlefield scenarios can be displayed and explored in real-time with minimal clutter. The user may also move units around by direct manipulation, define paths, create or delete hierarchical elements, and make other interactions. To strengthen the capacity for distributed simulations and for using sensor information from multiple sources, DIS capability has been integrated with the symbology for dynamic updates of position, direction and speed, and hierarchical structure. This paper will also discuss how the techniques used here can be applied to general (non-military) organizational structures.

Real-time weather data on terrain

This paper describes the visualization of 3D Doppler radar with global, with high-resolution terrain. This is the first time such data have been displayed together in a real-time environment. Associated data such as buildings and maps are displayed along with the weather data and the terrain. Requirements for effective 3D visualization for weather forecasting are identified. The application presented in this paper meets most of these requirements. In particular the application provides end-to-end real-time capability, integrated browsing and analysis, and integration of relevant data in a combined visualization. The last capability will grow in importance as researchers develop sophisticated models of storm development that yield rules for how storms behave in the presence of hills or mountains and other features.

Organization and simplification of high-resolution 3D city facades

This paper describes an approach for the organization and simplification of high-resolution geometry and imagery data for 3D buildings for interactive city navigation. At the highest level of organization, building data are inserted into a global hierarchy that supports the large-scale storage of cities around the world. This structure also provides fast access to the data suitable for interactive visualization. At this level the structure and simplification algorithms deal with city blocks. An associated latitude and longitude coordinate for each block is used to place it in the hierarchy. Each block is decomposed into building facades. A facade is a texture-mapped polygonal mesh representing one side of a city block. Therefore, a block typically contains four facades, but it may contain more. The facades are partitioned into relatively flat surfaces called faces. A texture-mapped polygonal mesh represents the building facades. By simplifying the faces first instead of the facades, the dominant characteristics of the building geometry are maintained. At the lowest level of detail, each face is simplified into a single texture-mapped polygon. An algorithm is presented for the simplification transition between the high- and low-detail representations of the faces. Other techniques for the simplification of entire blocks and even cities are discussed.