Post-doctoral offer: Radar textures for a unified modeling of sea and ground clutter


Hosted on the Palaiseau facility of ONERA, this post-doc aims at simulating radar land clutter from scatterers located on the 3D representation of a scene. The simulation should produce SAR images but also manage to recreate interferometric and polarimetric decorrelation from images that are simulated sequentially. The goal is to propose a model that encompasses both land and sea clutter, traditionally divided into two different simulating philosophies.


ONERA is a French public research organizationin the field of aeronautics and defense. ONERA’s missionsare developing and guidingresearch activities in the aerospace field as well as disseminating the results both in the aerospace industry and outside the aerospace field. With an annual budget of 236M€, ONERA has approximately 2000 employees in 7 centers throughout France.ONERA’s research covers all fields of aeronautics and space including electromagnetism and radar. Simulating EM response, radar signal and SAR images is a core skill of ONERA. This post-doc is specifically targeted at the simulation of SAR images.

Simulating a radar image of a complex scene, including targets in their surroundings (forests, rivers, roads…), requires an accurate representation of both types of elements. For targets, this is usually achieved via a meshed representation (CAD model). However, for surroundings, or radar clutter, this approach is either not feasible, due to a simple lack of knowledge, or extremely slow, due to the prohibitive mesh size. Instead, radar clutter is often simulatedwith statistical models.

For instance, to simulate radar clutter, it is common to use scatterers whose amplitude and/or phase are random variables drawn from a statistical distribution, with a power level dictated by the material properties (backscatter coefficient σ0). This approach is very fast but has some drawbacks. Notably, multichannel images (polarimetric or interferometric, temporal stack) have to be simulated all at once limiting the flexibility of the simulation. Thus, we wish to introduce some determinism in radar clutter simulation. To do so, we propose to use scatterers whose amplitude/or phase are given by deterministic and spatially correlated textures but modified by a random process. Besides, we propose to use the information contained in textures in conjunction with a scattering model (two-scale, small slope…[1,2]) to simulate the clutter consistently for different radar resolutions.

The goal of this post-doc is to develop such a random process, i.e., a model of perturbation of scatterers, which is compatible with textures and can simulate decorrelation in a consistent manner at different scales.Besides, up to now, radar clutter modeling has been almost strictly divided into two domains: sea and land clutter. There are several reasons forthis, namely scattering by the sea is akin to that of a rough surface while scattering by the ground and its cover also incorporates a volumic element. To establish such a model, we propose to build upon a point scattering propagation model for vegetationpreviously developed at ONERA [3]. Nevertheless, it is mandatory for a global simulation to have comparable degrees of accuracy for all models. This is why we seek to develop a unified model of sea and land clutter.

[1] F. Ticconi, L. Pulvirenti, et N. Pierdicca, « Models for scattering from rough surfaces » chap. 10, (2011).
[2] W J. Plant, « A two-scale model of short wind-generated waves and scatterometry », Journal of Geophysical, vol. 91, 10735-10749 (1986)
[3] J. Israel et al., « A propagation model for trees based on multiple scattering theory », EUCAP Conference (2016).

All informations are available in this document.