SBIR-STTR Award

General Reality Company proposes to develop an Automated Site Model Generation (ASMG) system capable of meeting DoD needs for a compact and automated "scene-to-simulator" image capture/processing device. The ASMG system uses a novel immersive camera which captures video over a near-spherical field-of-view to ensure full coverage of the terrain. 3D structure is extracted from the imagery using advanced motion factorization algorithms, assisted by an accurate 6DOF tracking system. Cultural features are recognized using a combination of object recognition algorithms and registration to DFAD/ITD data, and then inserted into an object-oriented hierarchical database where dynamic attributes are automatically and manually added. Finally, each cultural feature is converted to a set of polygons and textures, and output in standard formats such as Multigen OpenFlight, VRML, RADIUS, and DIS. To ensure the DoD and commercial users of ongoing improvements in the ASMG system, General Reality plans to develop the system in a modular fashion, and publish the architecture openly to support competitive improvements to each module. General Reality has assembled a strong team to carry out the proposed effort, including leading suppliers of each component technology.

General Reality Company (GRC) proposes to develop an Automated Site Model Generation (ASMG) System which automatically converts video imagery of real-world scenes into 3D object databases for close-combat virtual reality mission rehearsal simulations. During Phase I, GRC established the first known system approach capable of scaling to full-automation and near-real-time throughput. GRC's proprietary approach applies a novel combination of algorithms to automatically generate dense 3D depth maps of scenes imaged by a moving video camera. The depth maps are convened to individual 3D model objects using surface continuity algorithms coupled with geolocationbased indexing to a DMA database, and dynamic object attributes are generated by indexing to an attribute database. Output will be a hierarchical simulation database in Multigen fit format. The core algorithms will be further developed and demonstrated during the first year of the proposed project, followed by a six-month robustness enhancement effort. A Phase II option is planned to further enhance robustness and add specific features which reduce throughput time from 48 hours to 4 hours. In addition to supporting evolution of the Navy?s Tactical Training Environment Simulator from skills training roles to quick-turnaround mission rehearsal applications, the proposed effort has broad military applicability in close combat simulation, shipboard emergency simulation, target model generation, extreme-resolution terrain mapping, and combat engineering survey operations.