Simulation methodology using photographic terrain and rendered objects
Stephen J. Dow, Kevin Shelledy · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2001
We describe a simulation which has been developed over the last couple years for weapons trainers, in which a gunner can pan across a terrain scenario containing moving military vehicles. The terrain imagery is obtained by stitching together a panorama from photographs. Associated with this imagery are a wireframe model of the ground and a range image incorporating both the ground and other objects in the scene. This data allows military vehicles to be rendered at the correct scale and aspect for insertion in the scene, with occlusion by photographic scene objects as appropriate. Controls for contrast, brightness, and focus adjustment which are part of the weapons systems are included in the simulation, as well as rudimentary environmental effects. Achieving real-time performance with smooth user interaction on standard PC hardware posed some challenges. Our methodology assigns some of the processing steps and partial result storage to the graphics card, and others to the system CPU and memory. The methodology also makes frequent use of precomputed and previous fame partial results, so that most frames do not require all the processing step to be applied to the whole visible area. We present an overview of the frame construction methodology, and discuss in more detail the scheme for background precomputation of focus processing results.