Requirements for 3D Models to be Used in SAR Image Simulation

The simulation of Synthetic Aperture Radar (SAR) images plays an important role for many applications such as Automatic Target Recognition, damage detection or for the in-detail analysis of signatures found in real SAR images. The need for creating synthetic SAR images has increased with the advent of Convolutional Neural Networks (CNNs), since, in contrast to the Electro-Optical domain, only a few very specialized data sets of SAR images exist that could be used for training CNNs. Today, several academic and commercial SAR simulators exist. For all of these, the foundation for the simulation is a 3D model of the scene or object to be simulated. Such 3D models exist in large quantities and can be obtained for virtually any object imaginable. However, these models were not specifically designed for SAR simulation and thus the question arises whether they are suitable for the task. In this paper, several aspects of 3D models are highlighted that should be considered when assessing them for SAR simulation.