Modern radar simulation techniques can be used to generate artificial, yet fully realistic, radar sensor outputs such as Synthetic Aperture (SAR) and Real Beam (RB) ground map images. In general, when simulating the effect of an external channel, a numerical electromagnetic solver should have 3D models of the scene it is looking at. The creation of these scene models, as well as the rendering of radar images (using rasterization and/or ray tracing), parallels the developments and techniques of modern computer graphics (CG) systems. While CG techniques have mostly focused on the simulation of optical cameras, the physics of RF microwaves and radar signal processors means that simulation of a radar RF receiver requires unique, innovative techniques. Two fundamental differences between radar and optical systems are: information in the down-range dimension of radar data comes from range instead of perspective projection; and radar receiver data requires complex phase coherency, whereas conventional CG algorithms assume incoherent waves. In wide-area radar ground map images, features such as buildings and ground vehicles can be seen as strong returns from certain angles, but it is computationally impractical to use highly detailed mesh models for every scatterer in a scene. This paper reviews and adapts the modern CG techniques of Level-of-Detail (LOD) and Texture Mapping (including normal and tangent maps) to solve radar-specific simulation problems that increase the visual accuracy of the rendered images and decrease the time required to render. The improvements that are achieved with these techniques are of immense value for mission planning, radar scope interpretation training, and for the generation of synthetically generated training data for machine learning systems.
Keywords
ELECTRONIC WARFARE,FIDELITY,MACHINE LEARNING,MODELING,SIMULATIONS
Additional Keywords
Radar, computer graphics