KAISAR: Physics-based GPU-accelerated Realistic SAR Simulator

Abstract:
The enabling technology for Automatic Target Recognition (ATR) based on Artificial Intelligence is the availability of a database representative of realistic situations including targets and backgrounds. When applied to radar imaging the problem of collecting enough data of targets becomes extremely complicated and unaffordable. At MetaSensing we are tackling this problem by developing a physics-based SAR simulator, KAISAR, able to estimate the complex 3D RCS of a target and to model the SAR collection and processing stages to build a synthetic yet realistic database of target-background-images in a short time thanks to the use of GPU acceleration. This paper provides a description of the KAISAR, and examples of synthetic data generated.