Focus on... Sylvain Douté and image analysis of the planet Mars

on the April 18, 2019

Sylvain Douté is a researcher in Planetary Science at IPAG (Institute belonging to Grenoble Alpes University and CNRS). He obtained funding from the Grenoble Alpes Data Institute for his project "Massive analysis of multi-angular hyperspectral images of planet Mars by inverse regression of physical models".
Thanks to ground and space observations, Planetary Science generates numerous types of images that involve a large volume of data. Indeed, these images can have different dimensions according to acquisition angles and a multiplicity of colors belonging to the visible and infrared spectrum. Their interpretation allows determining the composition and the texture (i.e. grain size, transparency, and shape) of the materials present at the surface of planets.

With the support of the Grenoble Alpes Data Institute, Sylvain Douté (IPAG), Florence Forbes (INRIA) and their PhD student Benoit Kugler deal with the massive analysis of multi-angular hyperspectral images by the application of physical models. Capturing according to wavelength the sunlight reflected by different locations of the observed surface produces hyperspectral images. Now Physics tells us how the light reflection varies according to the acquisition geometry and to the medium properties on which the beam arrives. Thus Physical models have been produced that define the links between the physical properties of the surface and the observable.

The goal of the project is to create a method of mathematical interpretation that reverses the relationship between physical parameters and measurements. This method of inverse statistical regression aims at extracting quantitative information from spatial measurements. To the qualitative composition of a surface, quantitative methods add the abundance and the texture of the materials. Statistical methods have the advantage of revealing the quality of the extracted information by specifying its degree of uncertainty in the form of a probability distribution. The use of statistical regression techniques is also made necessary by the large volume of data to be processed. The observed surface is equivalent to all deserts on Earth and the detail of the sensors goes up to 25 cm.
Figure 1: Composite view of a portion of the Russell dune on Mars. Color encodes information on the composition, abundance, and texture of seasonal H2O and CO2 ices. During spring these deposits disappears by sublimation leading to areas with functioning dust plumes (dark blue spots).

Benoit Kugler is developing software that performs this inversion. It will soon be available in open source. An article will also be produced after the validation of the results. The software will notably be used in the context of the ExoMars program of the European Space Agency. A satellite has already been sent around Mars. A rover must leave in 2021 to explore the Martian surface. The landing site was chosen. Combined with other methods, the software will characterize the terrains of future exploration. Generated maps of properties will help planning the rover investigations.
Published on June 18, 2019