Current grid-based topographic products are not capable of representing near-vertical surfaces in the terrain such as ditches and washes to the fidelity required by modern scene based flight simulations. Torch proposes the development of a simple, efficient approach for the design of a toolset capable of identifying and generating high-fidelity near-vertical terrain features or using a-priori identification and integrating them into existing terrain topography at the high resolution required for an accurate thermal solution. This process incorporates modern image-processing techniques to identify and generate the high-fidelity terrain features and leverages massively parallel computational architectures to achieve this with a rapid, automated approach. The design is flexible both in software architecture, allowing multiple algorithms and methods of identifying and generating the terrain features, and hardware, allowing the usage of either high-density CPU or GPGPU architectures. By focusing on effective user interfaces and processes to minimize necessary labor, this allows the modeler to rapidly generate the necessary terrain models and integrate them into the thermal simulation and flight simulations for performance assessments. This process should allow integration into existing topographic software such as ERDAS Imagine, but is also relevant to many applications to the GeoSpatial community including feature detection and scene generation.
