Abstract
Karst aquifers, mainly limestone geological formations, represent a considerable challenge in the current context where many parts of the world are under water stress due to climate change. Although these aquifers are one of the world’s main freshwater sources, they are still underexploited due to a lack of precise knowledge of their geomorphology. Exploring them to collect data is a complex task because the networks formed in this type of environment are very extensive. In addition, sending divers is not ideal as they are limited in depth and the risks are significant due to the confinement and unpredictable structure. Eventually, the shipment of autonomous underwater vehicles would allow further exploration of the tunnels safely while acquiring data from sensors. However, using such vehicles in such an unpredictable environment without human supervision still presents many technical and scientific challenges.The objective of this thesis is therefore to propose 3D vision solutions adapted to gallery networks present in this type of environment. We thus propose two approaches based on the use of camera and a projector of conical shape.The first, prospective, is an active stereovision method using a system composed of two cameras and the projector, has been validated in simulation via the use of a gallery model.The second, which is the heart of this memory, is a structured light 3D reconstruction method using the combination of a camera and a cone-shaped projector.After extracting the contours, the points of these contours are reconstructed in 3D using the geometric constraint between the cone and the camera. This geometric constraint is expressed via the parameters of the headlamp cone, the estimation of which is made via a specific calibration.Various experiments have been carried out to validate this original approach. The first is carried out in a laboratory where a luminous contour projected on two orthogonal walls is reconstructed in 3D and then verified via tests of coplanarity and angulation between these two control walls. The second is carried out in an abandoned aqueduct (without water) having the advantage of reproducing an environment close to that of a karst aquifer: narrow, without light and especially without water to continue our assessment. The validation of the results is done on tests of coplanarity of the walls of the aqueduct and on the estimation of the known distance between these walls. The third is done in the pool, the first step before the final test in karstic environment. The validation of the results is also carried out by tests of the coplanarity of the points on the ground and the estimation of the cylindrical shape of the pool and its radius.