Visual Analysis of Mars Terrain Properties

Abstract

[EN] Currently, Martian Rover autonomous navigation techniques are based on geometrically derived navigational maps. Although effective in elevation map construction, these techniques do not succeed in detecting non geometrical hazards such as sand traps and duricrust layers (thin brittle layers of lava originated rock). The purpose of this thesis is to present an initial study of physics-based reflectance terrain classification and to evaluate its potential as a complementary system to the geometrical techniques. Section (11.2) contains the final output of this thesis, Bidirectional Reflectance Distribution Function (BRDF) plots, which constitute the core product that would be used by the proposed additional navigation system. This work concludes that the application of a clustering technique to the BRDF plots, which identified the different cluster distributions associated to each rock/soil class, would allow the development of a new physics-based terrain classifier. A measured BRDF plot would then be processed by this system and it would give the user the potential rock/soil class with a confidence percentage