Study and development of calibration algorithms of multi-camera systems for precision agriculture robotics

The aim of this thesis is to propose new techniques in the areas of mobile robotics. In particular, we studied innovative data processing techniques devoted to the estimation of the relative pose of a multicamera system, which can be used also on mobile robots. The estimated pose is obtained thanks to heterogeneous sensors mounted on the platform. So, a method of extrinsic calibration is used applied to a multicamera system. Later, the already calibrated multicamera system will be used for both robot localization and for the analysis of the surrounding environment. In order to validate all the methods, we utilize algorithms applied in general for the multicamera system, but we have validated them with a minimum number of cameras, i.e. two. These cameras are moving in solidarity each other in space. Each camera captures, in time, a cloud of 3D points and estimates the trajectory travelled in space, through SLAM algorithms. Based on reconstructed trajectories and joint motion constraint, the extrinsic calibration algorithm is able to estimate the relative pose between the two. So, practically, we test the algorithms first in simulation and then through a validation with a real multicamera system consisting of two cameras. After trying a number of procedures, that did not guarantee an optimal result, we were able to demonstrate through the validation that the methods leads to good calibration results and that therefore, if we know the trajectory accomplished by a robot, we can estimate the relative pose of the two cameras.