SAR simulation tools are typically used in the design process of SAR systems in the form of parametric studies that consider topics like prediction of image quality parameters, testing of image reconstruction algorithms, motion errors along the synthetic aperture, etc. The recent operational experience of space-borne SAR systems with sub-meter resolution (e.g. SAR-Lupe, COSMOSkyMed, and TerraSAR-X) for reconnaissance purposes demonstrated the importance of the understanding of SAR-specific image effects, especially foreshortening and layover, as well as shadow characteristics for the interpretation of complex targets like airplanes and ships.
SAR simulation could be a key to supply image operators with possibilities that simplify image interpretation and assist in applications like signature analysis or recognition. Therefore, a novel simulation framework has been established. It tries to fulfill all the above demands to the highest possible degree to reflect reality. The modular and flexible structure of the simulator allows adjusting and expanding it for different tasks.
Every simulation starts with an as accurate representation of the reality as possible. This is achieved by detailed 3-D surface model data of the object of interest. In the case of Figure 1 the wireframe model data represents a Boeing 747.
After setting SAR acquisition parameters like the attitude and the radar system configuration, a fast ray-tracing method finely samples the object of interest. Our approach not only finds the prominent scattering effects of the object, but also localizes them in the SAR image plane and calculates their reflectivity through sophisticated and highly runtime-efficient algorithms. The visual result of this step is called a reflectivity map and is illustrated in Figure 2 for the chosen Boeing 747 airplane.
Such a reflectivity map can be assumed to be an ideal SAR image, because it lacks effects like finite bandwidth or side lobe effects. Additional to the prominent scatterers, also the clutter return and the shadow of the airplane are included into this reflectivity map.
The final step in the simulation process has the task to take this ideal SAR signature and process it to a more realistic one. This is done by fast image reconstruction techniques that reduce the bandwidth of the reflectivity map by using the impulse response of the considered SAR system. The resulting simulated SAR signature as would be expected from a high resolution spotlight mode TerraSAR-X system configuration is illustrated in Figure 3.
To improve signature interpretability, sophisticated functionality to analyze scattering effects has been built into SAREF. Firstly, prominent scatterers can be color-coded according to their effect’s cause. Figure4 shows the simulated signature along with the color-coded prominent scattering effects of the object. Single-bounce scatterers are in yellow, edges in orange, double-bounce in green and finally multi-bounce effects in light blue.
Additionally the position of the scattering centers can not only be extracted in 2-D in the SAR image plane, but also in 3-D along with a transparent overlay of the wireframe model data, as can be seen from Figure 5. The stippled line in light blue indicates the simulated radar sensor’s line of sight.
Last but not least, the cause of prominent scatterers can be analyzed in a very convenient way by showing their ray-path. This has been done for a multi-bounce effect in Figure 6 which shows that it is caused by triple reflection between the airplane’s fuselage and the smooth tarmac ground.
SAREF can be used not only for the analysis of signatures, but also for their recognition. The workflow is demonstrated in the following video, which shows the whole procedure on the example of an airplane signature on Frankfurt airport imaged by TerraSAR-X with the staring spotlight mode. The basic approach for object recognition is to use a special optical view of the 3-D model data – the so-called perspective view – that exactly coincides with single-bounce scatterers in the SAR image plane. The alignment of this optical view onto the image plane is done interactively by the user.
The same principle in signature recognition of aligning the optical image plane with the one of the SAR image can also be used in larger scale for the whole scenery. Through this, the interactive alignment step by the user can be omitted, because the transformation between the image planes can be calculated automatically from the respective geographic information. In the following Figure 7 a SAR image of the harbor around Oslo is shown along with an aligned optical image. The optical image exactly fits onto signatures like buildings, because in the transformation process a high resolution digital elevation model (DEM) has been used (which also contains elevated structures like buildings). Optical and DEM image data is used by courtesy of the DLR Institute for Optical Systems.
The following video shows a further very powerful example of application. In the video a near real-time layover-shadow analysis is performed for the Oslo scene shown above. In combination with a contact analysis through orbit propagation (see system engineering research group) this functionality could be used for image acquisition planning.