Large object localization starting from 3D point clouds for defect analysis and detection

The latest automation trends are moving robotics away from traditional structured and predictable environments, to unstructured environments characterized by a higher degree of uncertainty and randomness. This thesis proposes the use of a 6D pose estimator based on point clouds in order to define the pose (position and orientation) of a large object in an unstructured environment. The proposed approach is based on the knowledge of the object's point cloud, obtainable from its CAD model, and on a real-time 3D scan of the environment in which the object is located. Through an extensive test phase, carried out both in a simulated and in a real life scenario, the performance of this approach was evaluated through the use of various properly defined KPIs. The fields of application of the approach are numerous. In the case covered by the thesis, the sector of interest was that of defect analysis and detection.