Résumé
Karst aquifers represent one of the most important freshwater resources in the world. However, these groundwaters are widely underexploited due to the complexity of conduits. In order to exploit these water resources, it is necessary to gather knowledge of their geometry. This paper presents a vision-based 3D reconstruction method using a stereo camera and a light projector to illuminate this confined environment. The originality of the approach is to leverage the projected halo of light on the environment surface for dense and robust stereo matching. We validate our approach through experiments in a simulated environments and demonstrate that contour reconstruction is suitable for this kind of environment.