Leveraging Gaming Technology To Generate High Resolution Geo-typical Terrain From Low Resolution Geo-specific Data For Varying Military Applications

Preparing synthetic or virtual 3D environments for use in military simulation can sometimes be seen as a ‘black art’. To achieve the required geospecific correlated terrain output, the process requires the correct coordinate system, the right projection/datum, access to digital elevation data and satellite imagery of sufficient resolution for the required application. The latter of these – access to high resolution satellite imagery – can be problematic on two fronts, namely cost (if available), or simply not being available from commercial sources. Add to this the additional requirement for intimate knowledge of several (often very expensive) disparate terrain software applications– let alone the target application and the path to preparing synthetic natural environments can seem very onerous. This paper explores the use of software intended for the gaming community to fast track the development of 3D environments for use in military simulation. Specifically, the following question is examined: For a given region of interest, where terrain data a. May not be available commercially b. May be available commercially, but is prohibitively expensive c. Is available open source, but is sparse / low resolution Is it possible to generate higher (yet configurable) resolution geo-typical terrain utilizing sparse / low resolution geo-specific terrain data which is available open source? The context for answering the above is a part of a wider technology development undertaken by Boeing’s Analysis, Modelling, Simulation and Experimentation international organization, this activity being undertaken from the Boeing Defence Australia’s Systems Analysis Laboratory. 1. INCENTIVE / RATIONALE A number of key reasons exist for investigating the use of gaming tools / methods as a means for procedurally “uprezing” both elevation and terrain texture data from readily available open source data. 1.1 Cost Firstly, the prohibitively expensive exercise of acquiring high resolution, high fidelity terrain data. In some instances this can run into the hundreds of thousands of dollars. If a means can be found for achieving a similar end result suitable for simulation requirements at a fraction of the cost, it is worth pursuing. Secondly, the huge outlay for specialized software used for manipulating terrain data. Again, if gaming technology and tools can be successfully leveraged to create a very close facsimile to their more expensive counterparts, the outcome is a plus. 1.2 Availability Another incentive may be lack of terrain data availability. For a given region of interest, high resolution imagery / elevation data may literally not be available. 1.3 Perceived Value Of Investigation Given the above considerations regarding cost and availability, this paper attributes a high value to finding a low cost alternative to the traditional means of acquiring high resolution terrain data. 2. TASK SCOPE The desired outcome examined in this paper is the ability to create high resolution, notionally accurate terrain data; in other words, a close facsimile of actual, real-world terrain regions. The task of creating such notionally accurate terrain data (as opposed to completely fictitious terrain) of sufficient detail for effective use in simulation presents a number of hurdles for the would be 3D environment developer, particularly where the end result must bear reasonably close scrutiny. In summary, the solution comes down to procedurally generating correlated terrain and imagery from low resolution data, essentially using the low resolution data as a starting point / reference and ending with a high resolution terrain data end result. That said, it is important to clarify up front what is and what isn’t in scope with respect to the aims of achieving the stated end goal. 2.1 Not in Scope In this paper, the term ‘real’ or ‘real-world’ with respect to terrain generation is not used to mean photo-realistic renderings. Regarding absolute fidelity, the generation of terrain data accurate to the last millimeter (geospecific) is not the concern of this paper. In fact, this paper seeks to explore alternate options to accommodate when such data is not available or is prohibitively expensive. Generating foliage, rocks etc: the generation of visual terrain enhancements such as trees, grass, rocks, boulders etc. is not the concern of this paper. 2.2 In Scope ‘Real’ or ‘real-world’ is being used to denote a specific terrain region with corresponding lat/lon coordinates. The aimed for generated terrain fidelity is a high degree of notional accuracy for the given region of interest, retaining correct lat/lon coordinate details with respect to the dominant terrain dimensions and features. The targeted end result is to output high resolution terrain data, allowing for an initial 1km sampled terrain elevation data to be re-sampled to 30 meters, and an even finer resolution for terrain texturing. Regarding procedural terrain synthesis, mention of available methods for increasing terrain geometry (elevation) resolution is made, including the use of fractal based interpolation as well as erosion algorithms. 2.3 Additional Considerations The following is a list (not exhaustive) of further factors taken into account for arriving at an investigation framework. 2.3.1 Computer Generated Forces For the purpose of creating terrain for use in military simulation, one important consideration is to ensure that any generated terrain output is also available for use in CGF (Computer Generated Forces) software. Examples of CGF software: • VR-Forces (http://www.mak.com/) [1] • STAGE (http://www.presagis.com) [2] • OneSAF (http://www.onesaf.net/community/) [3] 2.3.2 Communications Propagation Modeling Communications propagation modeling software (incorporating terrain effects such as occlusion) is another area which benefits from having access to highly detailed terrain elevation data. Examples of communications modeling software: • QualNet (http://www.scalable-networks.com/) [4] • OpNet (http://www.opnet.com/) [5] 2.3.3 Real-time Vs. Pre-processed Another important consideration is whether the process for increasing terrain resolution (e.g. via fractalisation) should be done prior to simulation (pre-process) or during simulation (real-time). For multiple participants in a simulation, all participants require identical LOD (Level Of Detail) and LOS (Line Of Sight) processing, even where disparate software is being used. In the instance of “uprezing” terrain details in real-time, the process would need to be identically implemented on each “client”. There is also the computational overhead to consider: “uprezing” elevation data during run-time is a considerable additional CPU overhead. Likewise, the calculation of path matrixes over terrain for AI is computationally intensive, far better implemented in the form of path lookup tables as a pre-simulation task. While it is possible to argue for a server/client implementation of the above utilizing highly parallel processing, at time of writing there appears to no viable candidate on the market to step up to the plate. Therefore this paper will focus on “pre-processing” as the strategy for terrain data “uprezing”. However, it is important to acknowledge that preprocessing does carry a couple of overheads. One is the potential size of the generated terrain database, which for a large area encompassing thousands of square kilometers can be considerable. The second problem area has to do with lack of agility, i.e. the length of time it takes to develop low resolution terrain data input into a useable, high resolution end result. 2.3.4 Tools This paper concerns itself with exploring the use of COTS gaming tools to fast track the development of 3D terrain for use in military simulation. 2.4 Investigation Framework Given the above considerations regarding requirements and scope, this paper will: • Explore the generation of high resolution geotypical terrain data from a low resolution, geospecific, open source starting point. • Ensure that the resulting terrain can be used in (current) CGF environments. • Select ‘pre-processing’ as the strategy for terrain “uprezing” • Restrict the investigation to COTS gaming tools. 3. EXAMPLES WITH LIMITATIONS The following is a software tools line-up. These tools are outstanding in their own right, but have limitations with respect to the current framework of investigation. 3.1 Blueberry 3D “... is a Procedural geometry engine allowing rapid development and visualization of realtime 3D databases in high resolution” (from website www.blueberry3d.com/) . This is an excellent example of real-time, dynamic terrain fractalisation and visualization. However, the tool does not fit within the current framework of investigation. 3.2 Terragen, Vue xStream These two software packages specialize in terrain rendering, and are also able to import heightfields and terrain imagery to generate (and export) elevation data. In addition, both are able to increase terrain elevation/imagery detail. Unfortunately, neither package has geo-spatial awareness. • Terragen (http://www.planetside.co.uk/terragen/) [7] • VUE (http://www.e-onsoftware.com/products/) [8] 3.3 Generalist 3D Software Generalist 3D software abounds both commercially as well as in open source form. Examples of Generalist 3D software: • Lightwave 3D (http://www.newtek.com/) [9] • Blender 3D (http://www.blender.org/) [10] • 3D Studio Max (www.autodesk.com) [11] • Maya (www.autodesk.com) [12] • XSI (www.autodesk.com) [13] Generalist 3D software excels in the area of texturing and manipulating 3D geometry, and usually allows for terrain data to be imported and applied to a mesh object. Additionally, further subdividing a mesh is native to most 3D modeling packages. However, while it is possible to import terrain data, the usual intent for generalist 3D software is a visual only output – i.e. nice renders. As a result, imported terrain data is used with a view to geometry (mesh) displacement and/or texturing only, and any geo-spatial data is discarded as being irrelevant. 4. TOOLS FOR CASE STUDY There is more than one software path for generating correlated terrain data from lower resolution input. The two case studies, however, will focus on only one software application, but acknowledge that other tools/techniques also have viability. 4.1 Grome The product “Grome” (Quad Software, http://www.quadsoftware.com) [14] is used for the case studies, a terrain creation/editing package. Able to create both fictitious terrain as well as import correlated terrain data (DTED), the software allows for the imported elevation/imagery to be increased in detail, and for the new, higher resolution data to be exported back out into DTED/GeoTIFF format. 4.2 Global Mapper In addition, the correlated terrain editor “Global Mapper” (Global Mapper Software, http://www.globalmapper.com/) [15] is utilized to aid the process of downloading open source data, and for validating the results generated by Grome. Global Mapper is not a necessitated part of the total tool chain, but serves as a utility of convenience. 5. CASE STUDY #1 Case Study #1 is a brief introduction to Grome and focuses on stepping the software through its stated terrain importing, texturing and exporting capabilities. 5.1 Description of Process For the purpose of demonstrating the process to proof of concept level (only), the region of Christmas Island was selected as “the region of interest”. The corresponding open source terrain data was downloaded into Global Mapper (elevation), and then saved out in DTED format for importing into Grome. 5.1.1 Terrain Texturing Once the elevation data was imported into Grome, an array of texturing tools and techniques were utilized to “dress” the terrain surface. Techniques can include leveraging the (downloadable) terrain surface imagery, as well as building up several texture “layers” based on altitude, slope angle, terrain facing direction etc. To further emulate the way terrain imagery is often available to the end user, it is also possible to have terrain “shadows” included as yet another texture layer. 5.1.2 Terrain Geometry Subdivision Available within Grome are a number means for synthesizing additional terrain detail, both with respect to interpolation as well as erosion. For interpolation, this includes fractalizing midpoints to create believable additional geometry. For purposes of creating further fine terrain details, the use of both thermal as well as fluid erosion models is available for application, the same being able to be restricted on the basis of “layering” as per texturing described above. The ability for Grome to seamlessly integrate elevation data of varying resolutions into one high resolution output is likewise a huge plus for tackling the objectives outlined elsewhere in this paper. 5.2 Open Source, Low Resolution Geo-specific Terrain Import into Global Mapper Global Mapper has an option for locating both commercial and open source terrain data. Figure 1. shows the dialogue window for accessing SRTM worldwide elevation data. Figure 1. GM: sourcing online elevation data. For the case study, the Christmas Island region (Figure 2.) was selected and the elevation data downloaded into Global Mapper. Figure 2. GM: Christmas Island elevation data Primarily for reference purposes, terrain imagery can also be downloaded into Global Mapper. Figure 3 shows Landsat7 Global Imagery being downloaded. Figure 3. GM: sourcing satellite imagery online 5.3 Importing Terrain Data into Grome Software The process for importing elevation data into Grome is straightforward: the software can read DTED files directly. Figure 4 shows the Christmas Island elevation data being imported. Figure 4. Import parameters and imported elevation data 5.4 Application of Texture Layer Tools in Grome The detailed procedure for applying texture layers in Grome is beyond the scope of this paper. Figure 5. shows the Christmas Island terrain textured to a sufficient level of detail to demonstrate the concept. Figure 5. Grome: Texture layering to proof of concept level. 5.5 Exporting High Resolution Geo-typical Terrain Data (elevation & imagery) from Grome Once the terrain data has been sufficiently textured and detailed, exporting the results directly to a higher resolution is a simple process. Figure 6. shows the geodata export dialogue windows. Figure 7. shows the resulting higher resolution Christmas Island data loaded into Global Mapper to validate the outcome. Figure 6. Grome: Exporting “uprezed” terrain elevation data plus high resolution terrain imagery Figure 7. Global Mapper: Grome’s high resolution Christmas Island terrain data loaded 6. CASE STUDY #2 Case Study #2 narrows the focus to demonstrating how terrain elevation data can be significantly increased in both resolution and definition. 6.1 Description of Process The region of Kieta was selected, and terrain data sampled at 1km (DTED 0) was imported into Grome. Figure 8. Global Mapper: DTED 0 of Kieta region, low resolution elevation data for importing into Grome 6.1.1 Terrain Texturing The texturing tools available in Grome were not invoked for this second case study. 6.1.2 Terrain Geometry Subdivision At time of importing the low resolution Kieta data, the default import sampling rate was increased from 1 km (DTED 0) to 30 meters (DTED 2). By increasing the resolution at time of import, Grome is able to apply erosion models to the imported terrain more effectively. Figure 9a. Grome: default import sampling increased from DTED 0 to DTED 2. Figure 9b. Grome: imported terrain region 6.1.3 Terrain Geometry Erosion In order to create increased definition on the surface of the imported terrain data, Grome’s terrain erosion feature was invoked. Specifically, the ‘Flow’ option of the Fluid Erosion model was used. Figure 10. Grome: Fluid Erosion model, ‘Flow’ option 6.1.4 Terrain Geometry Evaluation The resulting terrain was exported from Grome at DTED 2 resolution and loaded into Global Mapper for evaluation and comparison. The following images are a compare and contrast of the before and after increase of terrain resolution and definition using Grome. Figure 10a. DTED 0 Kieta source elevation data Figure 10b. DTED 0 Kieta source elevation data Figure 11a. DTED 2 Kieta, Grome erosion and export Figure 11b. DTED 2 Kieta, Grome erosion and export Figure 12a. DTED 1 Kieta, SRTM download Figure 12b. DTED 1 Kieta, SRTM download 7. FINDINGS The case studies demonstrate that the concept can be made to work: elevation data of limited detail can be enhanced with respect to resolution and definition, a high resolution texture “skin” for the terrain can be fast tracked with the right toolset, and the result can be exported for immediate usage to a wide variety of geospatially aware simulation applications. The use of Grome as a tool for the purpose of demonstrating that a comparatively low resolution, geospecific terrain source can be leveraged via COTS gaming technology to create high resolution, high definition geotypical data with a high degree of notional accuracy demonstrates the viability of the aims of this paper. 8. WAY FORWARDThe next step forward in evolving this technique wouldbe to develop a much larger terrain region to a highdegree of detail. Such a step could shake out any quirksor shortcomings in the software, as well as suggestrefinements with respect to workflow and toolsets.Identifying areas where terrain development can befurther fast-tracked through automation and/or scriptingwould be worth investigating as well. REFERENCESIn this paper reference is made to many non-Boeingproducts. Any comments regarding their utility arestrictly limited to the scope of this paper, and are notgeneral commentary on the quality of products or theirutility for other purposes (e.g. the creation of gamingassets). [1] VR-Forces, CGF Software, (http://www.mak.com/) [2] STAGE, CGF Software, (http://www.presagis.com) [3] OneSAF, CGF Software,(http://www.onesaf.net/community/) [4] QualNet, Communications Propagation ModelingSoftware, (http://www.scalable-networks.com/) [5] OpNet, Communications Propagation ModelingSoftware, (http://www.opnet.com/) [6] Blueberry 3D, Realtime Terrain VisualizationSoftware, (www.blueberry3d.com/) [7] Terragen, 3D Terrain Rendering Software,(http://www.planetside.co.uk/terragen/) [8] VUE, 3D Terrain Rendering Software,(http://www.e-onsoftware.com/products/) [9] Lightwave 3D, Modeling and Animation software,(http://www.newtek.com/) [10] Blender 3D, Modeling and Animation software,(http://www.blender.org/) [11] 3D Studio Max, Modeling and Animationsoftware, (www.autodesk.com) [12] Maya, Modeling and Animation software,(www.autodesk.com) [13] XSI, Modeling and Animation software,(www.autodesk.com) [14] Grome, Terrain Creation/Editing Software,(http://www.quadsoftware.com), Quad Software[15] Global Mapper GIS tool,(www.globalmapper.com), Global Mapper PLC. ACKNOWLEDGEMENTSSpecial thanks and acknowledgements to the followingpeople, who have provided encouragement as well aspractical support:Craig Pepper, Tech Team Lead, Systems AnalysisLaboratory, Boeing Defence Australia.Anthony Nixon, Model & Simulation Architect,Systems Analysis Laboratory, Boeing DefenceAustralia.Shane Arnott, International Deputy Director and ChiefTechnologist, Analysis, Modelling, Simulation andExperimentation, The Boeing Company