ETD

Archivio digitale delle tesi discusse presso l'Università di Pisa

Tesi etd-10272022-181250


Tipo di tesi
Tesi di laurea magistrale
Autore
BERNACCHI, LUCREZIA
URN
etd-10272022-181250
Titolo
Vision-based Underwater Object Detection and Autonomous Manipulation
Dipartimento
INGEGNERIA DELL'INFORMAZIONE
Corso di studi
INGEGNERIA ROBOTICA E DELL'AUTOMAZIONE
Relatori
relatore Prof. Costanzi, Riccardo
relatore Prof. Varagnolo, Damiano
correlatore Basso, Erlend Andreas
Parole chiave
  • underwater object detection
  • autonomous manipulation
Data inizio appello
24/11/2022
Consultabilità
Non consultabile
Data di rilascio
24/11/2092
Riassunto
This thesis aims to address autonomous manipulation tasks using a commercial, small size and low-cost underwater robot equipped with a gripper and a camera through which objects can be identified. The thesis finds concreteness in the METRICs EU project (Metrological Evaluation and
Testing of Robots in International Competitions) in particular in the RAMI competition (Robotics
for Asset Maintenance and Inspection) organised at the NATO STO CMRE (Science and Technology Organisation - Centre for Maritime Research and Experimentation) in La Spezia, Italy.
In the context of this competition, the thesis aims to perform a red valve and a pole-ring underwater detection with the purpose of performing manipulation tasks on the two objects, i.e., closing the valve and carrying the ring to the surface in complete autonomy.
This thesis proposes an object detection algorithm capable of detecting the two objects underwater
and, through the visual feedback, two manipulation strategies to perform the required tasks.
The proposed detection algorithm is based on two main concepts. The use of the a* channel of the L*a*b colour space to efficiently detect the presence of red objects underwater and perform an initial colour detection step. The use of convex envelopes to derive invariant features of the two objects, with respect to changes in orientation and position, and perform a last shape detection step in order to finally identify the valve and the ring. Using visual feedback, two autonomous manipulation strategies are then proposed to achieve the valve closure and the ring relocation. The proposed detection algorithm and the manipulation strategies has been tested in ROS-Gazebo framework at first and then in three real contexts: two artificial pools, at NTNU and at the Marine Cybernetics laboratory (Trondheim, Norway), and subsequently in the basin at the NATO centre (La Spezia, Italy) in the context of the RAMI competition.
File