Abstract
These thesis works are preliminary works to the construction of 3D-colour maps. They aim to solve the problem of combining LiDAR data and optical imagery acquired from a drone. Two prerequisites are identified. These consist, on the one hand, in characterizing the measurement errors of heterogeneous data from the sensors and, on the other hand, in geometrically aligning the latter.First, we propose the development of a LiDAR measurement uncertainty prediction model that takes into account the influence of the laser footprint. A new method without reference is introduced to validate this prediction model. A second method using a reference plane validates the adequacy of the use of the method without reference.In a second step, we propose a new method for calibrating the multi-sensor system consisting of a LiDAR, a camera, an inertial navigation system and a global satellite navigation system. The performance of this method is evaluated on synthetic and real data. It has the advantage of being fully automatic, does not require a calibration object or ground control point and can operate in either natural or urban environments. The flexibility of this method allows it to be implemented quickly before each acquisition.Finally, we propose a method to generate 3D-color maps in the form of colored point clouds. Our experiments show that geometric data alignment significantly improves the quality of 3D-color maps. If we look more closely at these 3D-colour maps, there are still colorization errors due mainly to the failure to take into account measurement errors. The use of the proposed LiDAR measurement uncertainty prediction model in the construction of 3D-color maps would therefore be the logical continuation of this work.