The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

Research Paper #615

Title:Vision Guided Part Alignment with Degraded Data
Date:Jan 1993
Presented:Presented at SICICA'92
Abstract:We investigate the possibility of controlling robot motion on the basis of features of a grasped part. A system was developed capable of inserting prisms of arbitrary cross-section into correspondingly shaped holes (the child's peg-in-hole problem). 2D vision processing reduces polygon approximations to both peg and hole cross-sections. A novel "shape attractor" algorithm is then used to find the motion in the image plane which brings these two sets of edges into alignment, and this in turn is converted to (approximate) robot motion, and the process is iterated. The system is self-calibrating and tolerant of degraded image data.

[Search These Pages] [DAI Home Page] [Comment]