Vision systems are more and more employed in robotic applications to perceive the surrounding environment and to make decisions accordingly. Despite this, the 3D reconstruction is a non-trivial problem for thin deformable linear objects. The cameras on the market that allow the 3D modelling of very small objects are bulky and expensive, and as a consequence cannot be easily integrated with a robotic system. This paper proposes a stereo vision system completely integrable into the robot end-effector, composed of two low-cost off-the-shelf endoscopic cameras with the aim of detecting tiny wires in the scene and identifying their location and diameter. The proposed method splices state-of-the-art vision algorithms applied to macro cameras obtaining, for wires with a diameter of few millimeters or less than a millimeter, a diameter estimation error lower than 10% and a location estimation error lower than 3%. The presented approach can be generalized to different types of endoscopic cameras and thin target objects.

Endoscopic Stereo Vision for Robotic 3D Detection of Thin Wire Features

Laudante G.;Mirto M.;Pennacchio O.;Pirozzi S.
2024

Abstract

Vision systems are more and more employed in robotic applications to perceive the surrounding environment and to make decisions accordingly. Despite this, the 3D reconstruction is a non-trivial problem for thin deformable linear objects. The cameras on the market that allow the 3D modelling of very small objects are bulky and expensive, and as a consequence cannot be easily integrated with a robotic system. This paper proposes a stereo vision system completely integrable into the robot end-effector, composed of two low-cost off-the-shelf endoscopic cameras with the aim of detecting tiny wires in the scene and identifying their location and diameter. The proposed method splices state-of-the-art vision algorithms applied to macro cameras obtaining, for wires with a diameter of few millimeters or less than a millimeter, a diameter estimation error lower than 10% and a location estimation error lower than 3%. The presented approach can be generalized to different types of endoscopic cameras and thin target objects.
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11591/547692
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? ND
social impact